linux-stable/scripts/mod/modpost.c

2294 lines
57 KiB
C
Raw Permalink Normal View History

/* Postprocess module symbol versions
*
* Copyright 2003 Kai Germaschewski
* Copyright 2002-2004 Rusty Russell, IBM Corporation
* Copyright 2006-2008 Sam Ravnborg
* Based in part on module-init-tools/depmod.c,file2alias
*
* This software may be used and distributed according to the terms
* of the GNU General Public License, incorporated herein by reference.
*
* Usage: modpost vmlinux module1.o module2.o ...
*/
#define _GNU_SOURCE
#include <elf.h>
#include <fnmatch.h>
#include <stdio.h>
#include <ctype.h>
#include <string.h>
#include <limits.h>
#include <stdbool.h>
modpost: Optionally ignore secondary errors seen if a single module build fails Commit ea4054a23 (modpost: handle huge numbers of modules) added support for building a large number of modules. Unfortunately, the commit changed the semantics of the makefile: Instead of passing only existing object files to modpost, make now passes all expected object files. If make was started with option -i, this results in a modpost error if a single file failed to build. Example with the current btrfs build falure on m68k: fs/btrfs/btrfs.o: No such file or directory make[1]: [__modpost] Error 1 (ignored) This error is followed by lots of errors such as: m68k-linux-gcc: error: arch/m68k/emu/nfcon.mod.c: No such file or directory m68k-linux-gcc: fatal error: no input files compilation terminated. make[1]: [arch/m68k/emu/nfcon.mod.o] Error 1 (ignored) This doesn't matter much for normal builds, but it is annoying for builds started with "make -i" due to the large number of secondary errors. Those errors unnececessarily clog any error log and make it difficult to find the real errors in the build. Fix the problem by adding a new parameter '-n' to modpost. If this parameter is specified, modpost reports but ignores missing object files. With this patch, error output from above problem is (with make -i): m68k-linux-ld: cannot find fs/btrfs/ioctl.o: No such file or directory make[2]: [fs/btrfs/btrfs.o] Error 1 (ignored) ... fs/btrfs/btrfs.o: No such file or directory (ignored) Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Michael Marek <mmarek@suse.cz> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
2013-09-23 05:53:54 +00:00
#include <errno.h>
#include "modpost.h"
#include "../../include/linux/license.h"
static bool module_enabled;
/* Are we using CONFIG_MODVERSIONS? */
static bool modversions;
/* Is CONFIG_MODULE_SRCVERSION_ALL set? */
static bool all_versions;
/* If we are modposting external module set to 1 */
static bool external_module;
/* Only warn about unresolved symbols */
static bool warn_unresolved;
modpost: change the license of EXPORT_SYMBOL to bool type There were more EXPORT_SYMBOL types in the past. The following commits removed unused ones. - f1c3d73e973c ("module: remove EXPORT_SYMBOL_GPL_FUTURE") - 367948220fce ("module: remove EXPORT_UNUSED_SYMBOL*") There are 3 remaining in enum export, but export_unknown does not make any sense because we never expect such a situation like "we do not know how it was exported". If the symbol name starts with "__ksymtab_", but the section name does not start with "___ksymtab+" or "___ksymtab_gpl+", it is not an exported symbol. It occurs when a variable starting with "__ksymtab_" is directly defined: int __ksymtab_foo; Presumably, there is no practical issue for using such a weird variable name (but there is no good reason for doing so, either). Anyway, that is not an exported symbol. Setting export_unknown is not the right thing to do. Do not call sym_add_exported() in this case. With pointless export_unknown removed, the export type finally becomes boolean (either EXPORT_SYMBOL or EXPORT_SYMBOL_GPL). I renamed the field name to is_gpl_only. EXPORT_SYMBOL_GPL sets it true. Only GPL-compatible modules can use it. I removed the orphan comment, "How a symbol is exported", which is unrelated to sec_mismatch_count. It is about enum export. See commit bd5cbcedf446 ("kbuild: export-type enhancement to modpost.c") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org>
2022-05-08 19:06:19 +00:00
static int sec_mismatch_count;
static bool sec_mismatch_warn_only = true;
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
/* Trim EXPORT_SYMBOLs that are unused by in-tree modules */
static bool trim_unused_exports;
modpost: Optionally ignore secondary errors seen if a single module build fails Commit ea4054a23 (modpost: handle huge numbers of modules) added support for building a large number of modules. Unfortunately, the commit changed the semantics of the makefile: Instead of passing only existing object files to modpost, make now passes all expected object files. If make was started with option -i, this results in a modpost error if a single file failed to build. Example with the current btrfs build falure on m68k: fs/btrfs/btrfs.o: No such file or directory make[1]: [__modpost] Error 1 (ignored) This error is followed by lots of errors such as: m68k-linux-gcc: error: arch/m68k/emu/nfcon.mod.c: No such file or directory m68k-linux-gcc: fatal error: no input files compilation terminated. make[1]: [arch/m68k/emu/nfcon.mod.o] Error 1 (ignored) This doesn't matter much for normal builds, but it is annoying for builds started with "make -i" due to the large number of secondary errors. Those errors unnececessarily clog any error log and make it difficult to find the real errors in the build. Fix the problem by adding a new parameter '-n' to modpost. If this parameter is specified, modpost reports but ignores missing object files. With this patch, error output from above problem is (with make -i): m68k-linux-ld: cannot find fs/btrfs/ioctl.o: No such file or directory make[2]: [fs/btrfs/btrfs.o] Error 1 (ignored) ... fs/btrfs/btrfs.o: No such file or directory (ignored) Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Michael Marek <mmarek@suse.cz> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
2013-09-23 05:53:54 +00:00
/* ignore missing files */
static bool ignore_missing_files;
/* If set to 1, only warn (instead of error) about missing ns imports */
static bool allow_missing_ns_imports;
kbuild: add verbose option to Section mismatch reporting in modpost If the config option CONFIG_SECTION_MISMATCH is not set and we see a Section mismatch present the following to the user: modpost: Found 1 section mismatch(es). To see additional details select "Enable full Section mismatch analysis" in the Kernel Hacking menu (CONFIG_SECTION_MISMATCH). If the option CONFIG_SECTION_MISMATCH is selected then be verbose in the Section mismatch reporting from mdopost. Sample outputs: WARNING: o-x86_64/vmlinux.o(.text+0x7396): Section mismatch in reference from the function discover_ebda() to the variable .init.data:ebda_addr The function discover_ebda() references the variable __initdata ebda_addr. This is often because discover_ebda lacks a __initdata annotation or the annotation of ebda_addr is wrong. WARNING: o-x86_64/vmlinux.o(.data+0x74d58): Section mismatch in reference from the variable pci_serial_quirks to the function .devexit.text:pci_plx9050_exit() The variable pci_serial_quirks references the function __devexit pci_plx9050_exit() If the reference is valid then annotate the variable with __exit* (see linux/init.h) or name the variable: *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, WARNING: o-x86_64/vmlinux.o(__ksymtab+0x630): Section mismatch in reference from the variable __ksymtab_arch_register_cpu to the function .cpuinit.text:arch_register_cpu() The symbol arch_register_cpu is exported and annotated __cpuinit Fix this by removing the __cpuinit annotation of arch_register_cpu or drop the export. Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
2008-01-24 20:12:37 +00:00
static bool error_occurred;
static bool extra_warn;
kbuild: fix false-positive modpost warning when all symbols are trimmed Nathan reports that the mips defconfig emits the following warning: WARNING: modpost: Symbol info of vmlinux is missing. Unresolved symbol check will be entirely skipped. This false-positive happens when CONFIG_TRIM_UNUSED_KSYMS is enabled, but no CONFIG option is set to 'm'. Commit a0590473c5e6 ("nfs: fix PNFS_FLEXFILE_LAYOUT Kconfig default") turned the last 'm' into 'y' for the mips defconfig, and uncovered this issue. In this case, the module feature itself is enabled, but we have no module to build. As a result, CONFIG_TRIM_UNUSED_KSYMS drops all the instances of EXPORT_SYMBOL. Then, modpost wrongly assumes vmlinux is missing because vmlinux.symvers is empty. (As another false-positive case, you can create a module that does not use any symbol of vmlinux). The current behavior is to entirely suppress the unresolved symbol warnings when vmlinux is missing just because there are too many. I found the origin of this code in the historical git tree. [1] If this is a matter of noisiness, I think modpost can display the first 10 warnings, and the number of suppressed warnings at the end. You will get a bit noisier logs when you run 'make modules' without vmlinux, but such warnings are better to show because you never know the resulting modules are actually loadable or not. This commit changes the following: - If any of input *.symver files is missing, pass -w option to let the module build keep going with warnings instead of errors. - If there are too many (10+) unresolved symbol warnings, show only the first 10, and also the number of suppressed warnings. [1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=1cc0e0529569bf6a94f6d49770aa6d4b599d2c46 Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2021-03-25 18:54:11 +00:00
/*
* Cut off the warnings when there are too many. This typically occurs when
* vmlinux is missing. ('make modules' without building vmlinux.)
*/
#define MAX_UNRESOLVED_REPORTS 10
static unsigned int nr_unresolved;
/* In kernel, this size is defined in linux/module.h;
* here we use Elf_Addr instead of long for covering cross-compile
*/
#define MODULE_NAME_LEN (64 - sizeof(Elf_Addr))
void modpost_log(enum loglevel loglevel, const char *fmt, ...)
{
va_list arglist;
switch (loglevel) {
case LOG_WARN:
fprintf(stderr, "WARNING: ");
break;
case LOG_ERROR:
fprintf(stderr, "ERROR: ");
error_occurred = true;
break;
default: /* invalid loglevel, ignore */
break;
}
fprintf(stderr, "modpost: ");
va_start(arglist, fmt);
vfprintf(stderr, fmt, arglist);
va_end(arglist);
}
static inline bool strends(const char *str, const char *postfix)
{
if (strlen(str) < strlen(postfix))
return false;
return strcmp(str + strlen(str) - strlen(postfix), postfix) == 0;
}
void *do_nofail(void *ptr, const char *expr)
{
if (!ptr)
fatal("Memory allocation failure: %s.\n", expr);
return ptr;
}
char *read_text_file(const char *filename)
{
struct stat st;
size_t nbytes;
int fd;
char *buf;
fd = open(filename, O_RDONLY);
if (fd < 0) {
perror(filename);
exit(1);
}
if (fstat(fd, &st) < 0) {
perror(filename);
exit(1);
}
buf = NOFAIL(malloc(st.st_size + 1));
nbytes = st.st_size;
while (nbytes) {
ssize_t bytes_read;
bytes_read = read(fd, buf, nbytes);
if (bytes_read < 0) {
perror(filename);
exit(1);
}
nbytes -= bytes_read;
}
buf[st.st_size] = '\0';
close(fd);
return buf;
}
char *get_line(char **stringp)
{
char *orig = *stringp, *next;
/* do not return the unwanted extra line at EOF */
if (!orig || *orig == '\0')
return NULL;
/* don't use strsep here, it is not available everywhere */
next = strchr(orig, '\n');
if (next)
*next++ = '\0';
*stringp = next;
return orig;
}
/* A list of all modules we processed */
LIST_HEAD(modules);
static struct module *find_module(const char *modname)
{
struct module *mod;
list_for_each_entry(mod, &modules, list) {
if (strcmp(mod->name, modname) == 0)
return mod;
}
return NULL;
}
static struct module *new_module(const char *name, size_t namelen)
{
struct module *mod;
mod = NOFAIL(malloc(sizeof(*mod) + namelen + 1));
memset(mod, 0, sizeof(*mod));
INIT_LIST_HEAD(&mod->exported_symbols);
INIT_LIST_HEAD(&mod->unresolved_symbols);
INIT_LIST_HEAD(&mod->missing_namespaces);
INIT_LIST_HEAD(&mod->imported_namespaces);
memcpy(mod->name, name, namelen);
mod->name[namelen] = '\0';
mod->is_vmlinux = (strcmp(mod->name, "vmlinux") == 0);
/*
* Set mod->is_gpl_compatible to true by default. If MODULE_LICENSE()
* is missing, do not check the use for EXPORT_SYMBOL_GPL() becasue
* modpost will exit wiht error anyway.
*/
mod->is_gpl_compatible = true;
list_add_tail(&mod->list, &modules);
return mod;
}
/* A hash of all exported symbols,
* struct symbol is also used for lists of unresolved symbols */
#define SYMBOL_HASH_SIZE 1024
struct symbol {
struct symbol *next;
struct list_head list; /* link to module::exported_symbols or module::unresolved_symbols */
struct module *module;
modpost: fix broken sym->namespace for external module builds Currently, external module builds produce tons of false-positives: WARNING: module <mod> uses symbol <sym> from namespace <ns>, but does not import it. Here, the <ns> part shows a random string. When you build external modules, the symbol info of vmlinux and in-kernel modules are read from $(objtree)/Module.symvers, but read_dump() is buggy in multiple ways: [1] When the modpost is run for vmlinux and in-kernel modules, sym_extract_namespace() allocates memory for the namespace. On the other hand, read_dump() does not, then sym->namespace will point to somewhere in the line buffer of get_next_line(). The data in the buffer will be replaced soon, and sym->namespace will end up with pointing to unrelated data. As a result, check_exports() will show random strings in the warning messages. [2] When there is no namespace, sym_extract_namespace() returns NULL. On the other hand, read_dump() sets namespace to an empty string "". (but, it will be later replaced with unrelated data due to bug [1].) The check_exports() shows a warning unless exp->namespace is NULL, so every symbol read from read_dump() emits the warning, which is mostly false positive. To address [1], sym_add_exported() calls strdup() for s->namespace. The namespace from sym_extract_namespace() must be freed to avoid memory leak. For [2], I changed the if-conditional in check_exports(). This commit also fixes sym_add_exported() to set s->namespace correctly when the symbol is preloaded. Reviewed-by: Matthias Maennich <maennich@google.com> Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com> Signed-off-by: Jessica Yu <jeyu@kernel.org>
2019-10-03 07:58:22 +00:00
char *namespace;
unsigned int crc;
bool crc_valid;
bool weak;
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
bool is_func;
modpost: change the license of EXPORT_SYMBOL to bool type There were more EXPORT_SYMBOL types in the past. The following commits removed unused ones. - f1c3d73e973c ("module: remove EXPORT_SYMBOL_GPL_FUTURE") - 367948220fce ("module: remove EXPORT_UNUSED_SYMBOL*") There are 3 remaining in enum export, but export_unknown does not make any sense because we never expect such a situation like "we do not know how it was exported". If the symbol name starts with "__ksymtab_", but the section name does not start with "___ksymtab+" or "___ksymtab_gpl+", it is not an exported symbol. It occurs when a variable starting with "__ksymtab_" is directly defined: int __ksymtab_foo; Presumably, there is no practical issue for using such a weird variable name (but there is no good reason for doing so, either). Anyway, that is not an exported symbol. Setting export_unknown is not the right thing to do. Do not call sym_add_exported() in this case. With pointless export_unknown removed, the export type finally becomes boolean (either EXPORT_SYMBOL or EXPORT_SYMBOL_GPL). I renamed the field name to is_gpl_only. EXPORT_SYMBOL_GPL sets it true. Only GPL-compatible modules can use it. I removed the orphan comment, "How a symbol is exported", which is unrelated to sec_mismatch_count. It is about enum export. See commit bd5cbcedf446 ("kbuild: export-type enhancement to modpost.c") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org>
2022-05-08 19:06:19 +00:00
bool is_gpl_only; /* exported by EXPORT_SYMBOL_GPL */
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
bool used; /* there exists a user of this symbol */
modpost,fixdep: Replace zero-length array with flexible-array The current codebase makes use of the zero-length array language extension to the C90 standard, but the preferred mechanism to declare variable-length types such as these ones is a flexible array member[1][2], introduced in C99: struct foo { int stuff; struct boo array[]; }; By making use of the mechanism above, we will get a compiler warning in case the flexible array does not occur last in the structure, which will help us prevent some kind of undefined behavior bugs from being inadvertently introduced[3] to the codebase from now on. Also, notice that, dynamic memory allocations won't be affected by this change: "Flexible array members have incomplete type, and so the sizeof operator may not be applied. As a quirk of the original implementation of zero-length arrays, sizeof evaluates to zero."[1] sizeof(flexible-array-member) triggers a warning because flexible array members have incomplete type[1]. There are some instances of code in which the sizeof operator is being incorrectly/erroneously applied to zero-length arrays and the result is zero. Such instances may be hiding some bugs. So, this work (flexible-array member conversions) will also help to get completely rid of those sorts of issues. This issue was found with the help of Coccinelle. [1] https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html [2] https://github.com/KSPP/linux/issues/21 [3] commit 76497732932f ("cxgb3/l2t: Fix undefined behaviour") Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2020-05-07 18:56:01 +00:00
char name[];
};
static struct symbol *symbolhash[SYMBOL_HASH_SIZE];
/* This is based on the hash algorithm from gdbm, via tdb */
static inline unsigned int tdb_hash(const char *name)
{
unsigned value; /* Used to compute the hash value. */
unsigned i; /* Used to cycle through random values. */
/* Set the initial value from the key size. */
for (value = 0x238F13AF * strlen(name), i = 0; name[i]; i++)
value = (value + (((unsigned char *)name)[i] << (i*5 % 24)));
return (1103515243 * value + 12345);
}
/**
* Allocate a new symbols for use in the hash of exported symbols or
* the list of unresolved symbols per module
**/
static struct symbol *alloc_symbol(const char *name)
{
struct symbol *s = NOFAIL(malloc(sizeof(*s) + strlen(name) + 1));
memset(s, 0, sizeof(*s));
strcpy(s->name, name);
return s;
}
/* For the hash of exported symbols */
static void hash_add_symbol(struct symbol *sym)
{
unsigned int hash;
hash = tdb_hash(sym->name) % SYMBOL_HASH_SIZE;
sym->next = symbolhash[hash];
symbolhash[hash] = sym;
}
static void sym_add_unresolved(const char *name, struct module *mod, bool weak)
{
struct symbol *sym;
sym = alloc_symbol(name);
sym->weak = weak;
list_add_tail(&sym->list, &mod->unresolved_symbols);
}
static struct symbol *sym_find_with_module(const char *name, struct module *mod)
{
struct symbol *s;
/* For our purposes, .foo matches foo. PPC64 needs this. */
if (name[0] == '.')
name++;
for (s = symbolhash[tdb_hash(name) % SYMBOL_HASH_SIZE]; s; s = s->next) {
if (strcmp(s->name, name) == 0 && (!mod || s->module == mod))
return s;
}
return NULL;
}
static struct symbol *find_symbol(const char *name)
{
return sym_find_with_module(name, NULL);
}
struct namespace_list {
struct list_head list;
char namespace[];
};
static bool contains_namespace(struct list_head *head, const char *namespace)
{
struct namespace_list *list;
/*
* The default namespace is null string "", which is always implicitly
* contained.
*/
if (!namespace[0])
return true;
list_for_each_entry(list, head, list) {
if (!strcmp(list->namespace, namespace))
return true;
}
return false;
}
static void add_namespace(struct list_head *head, const char *namespace)
{
struct namespace_list *ns_entry;
if (!contains_namespace(head, namespace)) {
ns_entry = NOFAIL(malloc(sizeof(*ns_entry) +
strlen(namespace) + 1));
strcpy(ns_entry->namespace, namespace);
list_add_tail(&ns_entry->list, head);
}
}
static void *sym_get_data_by_offset(const struct elf_info *info,
unsigned int secindex, unsigned long offset)
{
Elf_Shdr *sechdr = &info->sechdrs[secindex];
return (void *)info->hdr + sechdr->sh_offset + offset;
}
void *sym_get_data(const struct elf_info *info, const Elf_Sym *sym)
{
return sym_get_data_by_offset(info, get_secindex(info, sym),
sym->st_value);
}
static const char *sech_name(const struct elf_info *info, Elf_Shdr *sechdr)
{
return sym_get_data_by_offset(info, info->secindex_strings,
sechdr->sh_name);
}
static const char *sec_name(const struct elf_info *info, unsigned int secindex)
{
/*
* If sym->st_shndx is a special section index, there is no
* corresponding section header.
* Return "" if the index is out of range of info->sechdrs[] array.
*/
if (secindex >= info->num_sections)
return "";
return sech_name(info, &info->sechdrs[secindex]);
}
#define strstarts(str, prefix) (strncmp(str, prefix, strlen(prefix)) == 0)
static struct symbol *sym_add_exported(const char *name, struct module *mod,
bool gpl_only, const char *namespace)
{
struct symbol *s = find_symbol(name);
if (s && (!external_module || s->module->is_vmlinux || s->module == mod)) {
error("%s: '%s' exported twice. Previous export was in %s%s\n",
mod->name, name, s->module->name,
s->module->is_vmlinux ? "" : ".ko");
}
s = alloc_symbol(name);
s->module = mod;
modpost: change the license of EXPORT_SYMBOL to bool type There were more EXPORT_SYMBOL types in the past. The following commits removed unused ones. - f1c3d73e973c ("module: remove EXPORT_SYMBOL_GPL_FUTURE") - 367948220fce ("module: remove EXPORT_UNUSED_SYMBOL*") There are 3 remaining in enum export, but export_unknown does not make any sense because we never expect such a situation like "we do not know how it was exported". If the symbol name starts with "__ksymtab_", but the section name does not start with "___ksymtab+" or "___ksymtab_gpl+", it is not an exported symbol. It occurs when a variable starting with "__ksymtab_" is directly defined: int __ksymtab_foo; Presumably, there is no practical issue for using such a weird variable name (but there is no good reason for doing so, either). Anyway, that is not an exported symbol. Setting export_unknown is not the right thing to do. Do not call sym_add_exported() in this case. With pointless export_unknown removed, the export type finally becomes boolean (either EXPORT_SYMBOL or EXPORT_SYMBOL_GPL). I renamed the field name to is_gpl_only. EXPORT_SYMBOL_GPL sets it true. Only GPL-compatible modules can use it. I removed the orphan comment, "How a symbol is exported", which is unrelated to sec_mismatch_count. It is about enum export. See commit bd5cbcedf446 ("kbuild: export-type enhancement to modpost.c") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org>
2022-05-08 19:06:19 +00:00
s->is_gpl_only = gpl_only;
s->namespace = NOFAIL(strdup(namespace));
list_add_tail(&s->list, &mod->exported_symbols);
hash_add_symbol(s);
return s;
}
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
static void sym_set_crc(struct symbol *sym, unsigned int crc)
{
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
sym->crc = crc;
sym->crc_valid = true;
}
static void *grab_file(const char *filename, size_t *size)
{
struct stat st;
void *map = MAP_FAILED;
int fd;
fd = open(filename, O_RDONLY);
if (fd < 0)
return NULL;
if (fstat(fd, &st))
goto failed;
*size = st.st_size;
map = mmap(NULL, *size, PROT_READ|PROT_WRITE, MAP_PRIVATE, fd, 0);
failed:
close(fd);
if (map == MAP_FAILED)
return NULL;
return map;
}
static void release_file(void *file, size_t size)
{
munmap(file, size);
}
static int parse_elf(struct elf_info *info, const char *filename)
{
unsigned int i;
Elf_Ehdr *hdr;
Elf_Shdr *sechdrs;
Elf_Sym *sym;
const char *secstrings;
unsigned int symtab_idx = ~0U, symtab_shndx_idx = ~0U;
hdr = grab_file(filename, &info->size);
if (!hdr) {
modpost: Optionally ignore secondary errors seen if a single module build fails Commit ea4054a23 (modpost: handle huge numbers of modules) added support for building a large number of modules. Unfortunately, the commit changed the semantics of the makefile: Instead of passing only existing object files to modpost, make now passes all expected object files. If make was started with option -i, this results in a modpost error if a single file failed to build. Example with the current btrfs build falure on m68k: fs/btrfs/btrfs.o: No such file or directory make[1]: [__modpost] Error 1 (ignored) This error is followed by lots of errors such as: m68k-linux-gcc: error: arch/m68k/emu/nfcon.mod.c: No such file or directory m68k-linux-gcc: fatal error: no input files compilation terminated. make[1]: [arch/m68k/emu/nfcon.mod.o] Error 1 (ignored) This doesn't matter much for normal builds, but it is annoying for builds started with "make -i" due to the large number of secondary errors. Those errors unnececessarily clog any error log and make it difficult to find the real errors in the build. Fix the problem by adding a new parameter '-n' to modpost. If this parameter is specified, modpost reports but ignores missing object files. With this patch, error output from above problem is (with make -i): m68k-linux-ld: cannot find fs/btrfs/ioctl.o: No such file or directory make[2]: [fs/btrfs/btrfs.o] Error 1 (ignored) ... fs/btrfs/btrfs.o: No such file or directory (ignored) Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Michael Marek <mmarek@suse.cz> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
2013-09-23 05:53:54 +00:00
if (ignore_missing_files) {
fprintf(stderr, "%s: %s (ignored)\n", filename,
strerror(errno));
return 0;
}
perror(filename);
exit(1);
}
info->hdr = hdr;
if (info->size < sizeof(*hdr)) {
/* file too small, assume this is an empty .o file */
return 0;
}
/* Is this a valid ELF file? */
if ((hdr->e_ident[EI_MAG0] != ELFMAG0) ||
(hdr->e_ident[EI_MAG1] != ELFMAG1) ||
(hdr->e_ident[EI_MAG2] != ELFMAG2) ||
(hdr->e_ident[EI_MAG3] != ELFMAG3)) {
/* Not an ELF file - silently ignore it */
return 0;
}
/* Fix endianness in ELF header */
hdr->e_type = TO_NATIVE(hdr->e_type);
hdr->e_machine = TO_NATIVE(hdr->e_machine);
hdr->e_version = TO_NATIVE(hdr->e_version);
hdr->e_entry = TO_NATIVE(hdr->e_entry);
hdr->e_phoff = TO_NATIVE(hdr->e_phoff);
hdr->e_shoff = TO_NATIVE(hdr->e_shoff);
hdr->e_flags = TO_NATIVE(hdr->e_flags);
hdr->e_ehsize = TO_NATIVE(hdr->e_ehsize);
hdr->e_phentsize = TO_NATIVE(hdr->e_phentsize);
hdr->e_phnum = TO_NATIVE(hdr->e_phnum);
hdr->e_shentsize = TO_NATIVE(hdr->e_shentsize);
hdr->e_shnum = TO_NATIVE(hdr->e_shnum);
hdr->e_shstrndx = TO_NATIVE(hdr->e_shstrndx);
sechdrs = (void *)hdr + hdr->e_shoff;
info->sechdrs = sechdrs;
/* modpost only works for relocatable objects */
if (hdr->e_type != ET_REL)
fatal("%s: not relocatable object.", filename);
/* Check if file offset is correct */
if (hdr->e_shoff > info->size)
fatal("section header offset=%lu in file '%s' is bigger than filesize=%zu\n",
(unsigned long)hdr->e_shoff, filename, info->size);
if (hdr->e_shnum == SHN_UNDEF) {
/*
* There are more than 64k sections,
* read count from .sh_size.
*/
info->num_sections = TO_NATIVE(sechdrs[0].sh_size);
}
else {
info->num_sections = hdr->e_shnum;
}
if (hdr->e_shstrndx == SHN_XINDEX) {
info->secindex_strings = TO_NATIVE(sechdrs[0].sh_link);
}
else {
info->secindex_strings = hdr->e_shstrndx;
}
/* Fix endianness in section headers */
for (i = 0; i < info->num_sections; i++) {
sechdrs[i].sh_name = TO_NATIVE(sechdrs[i].sh_name);
sechdrs[i].sh_type = TO_NATIVE(sechdrs[i].sh_type);
sechdrs[i].sh_flags = TO_NATIVE(sechdrs[i].sh_flags);
sechdrs[i].sh_addr = TO_NATIVE(sechdrs[i].sh_addr);
sechdrs[i].sh_offset = TO_NATIVE(sechdrs[i].sh_offset);
sechdrs[i].sh_size = TO_NATIVE(sechdrs[i].sh_size);
sechdrs[i].sh_link = TO_NATIVE(sechdrs[i].sh_link);
sechdrs[i].sh_info = TO_NATIVE(sechdrs[i].sh_info);
sechdrs[i].sh_addralign = TO_NATIVE(sechdrs[i].sh_addralign);
sechdrs[i].sh_entsize = TO_NATIVE(sechdrs[i].sh_entsize);
}
/* Find symbol table. */
secstrings = (void *)hdr + sechdrs[info->secindex_strings].sh_offset;
for (i = 1; i < info->num_sections; i++) {
const char *secname;
int nobits = sechdrs[i].sh_type == SHT_NOBITS;
if (!nobits && sechdrs[i].sh_offset > info->size)
fatal("%s is truncated. sechdrs[i].sh_offset=%lu > sizeof(*hrd)=%zu\n",
filename, (unsigned long)sechdrs[i].sh_offset,
sizeof(*hdr));
secname = secstrings + sechdrs[i].sh_name;
if (strcmp(secname, ".modinfo") == 0) {
if (nobits)
fatal("%s has NOBITS .modinfo\n", filename);
info->modinfo = (void *)hdr + sechdrs[i].sh_offset;
info->modinfo_len = sechdrs[i].sh_size;
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
} else if (!strcmp(secname, ".export_symbol")) {
info->export_symbol_secndx = i;
}
if (sechdrs[i].sh_type == SHT_SYMTAB) {
unsigned int sh_link_idx;
symtab_idx = i;
info->symtab_start = (void *)hdr +
sechdrs[i].sh_offset;
info->symtab_stop = (void *)hdr +
sechdrs[i].sh_offset + sechdrs[i].sh_size;
sh_link_idx = sechdrs[i].sh_link;
info->strtab = (void *)hdr +
sechdrs[sh_link_idx].sh_offset;
}
/* 32bit section no. table? ("more than 64k sections") */
if (sechdrs[i].sh_type == SHT_SYMTAB_SHNDX) {
symtab_shndx_idx = i;
info->symtab_shndx_start = (void *)hdr +
sechdrs[i].sh_offset;
info->symtab_shndx_stop = (void *)hdr +
sechdrs[i].sh_offset + sechdrs[i].sh_size;
}
}
if (!info->symtab_start)
fatal("%s has no symtab?\n", filename);
/* Fix endianness in symbols */
for (sym = info->symtab_start; sym < info->symtab_stop; sym++) {
sym->st_shndx = TO_NATIVE(sym->st_shndx);
sym->st_name = TO_NATIVE(sym->st_name);
sym->st_value = TO_NATIVE(sym->st_value);
sym->st_size = TO_NATIVE(sym->st_size);
}
if (symtab_shndx_idx != ~0U) {
Elf32_Word *p;
if (symtab_idx != sechdrs[symtab_shndx_idx].sh_link)
fatal("%s: SYMTAB_SHNDX has bad sh_link: %u!=%u\n",
filename, sechdrs[symtab_shndx_idx].sh_link,
symtab_idx);
/* Fix endianness */
for (p = info->symtab_shndx_start; p < info->symtab_shndx_stop;
p++)
*p = TO_NATIVE(*p);
}
symsearch_init(info);
return 1;
}
static void parse_elf_finish(struct elf_info *info)
{
symsearch_finish(info);
release_file(info->hdr, info->size);
}
static int ignore_undef_symbol(struct elf_info *info, const char *symname)
{
/* ignore __this_module, it will be resolved shortly */
if (strcmp(symname, "__this_module") == 0)
return 1;
/* ignore global offset table */
if (strcmp(symname, "_GLOBAL_OFFSET_TABLE_") == 0)
return 1;
if (info->hdr->e_machine == EM_PPC)
/* Special register function linked on all modules during final link of .ko */
if (strstarts(symname, "_restgpr_") ||
strstarts(symname, "_savegpr_") ||
strstarts(symname, "_rest32gpr_") ||
strstarts(symname, "_save32gpr_") ||
strstarts(symname, "_restvr_") ||
strstarts(symname, "_savevr_"))
return 1;
if (info->hdr->e_machine == EM_PPC64)
/* Special register function linked on all modules during final link of .ko */
if (strstarts(symname, "_restgpr0_") ||
strstarts(symname, "_savegpr0_") ||
strstarts(symname, "_restvr_") ||
strstarts(symname, "_savevr_") ||
strcmp(symname, ".TOC.") == 0)
return 1;
/* Do not ignore this symbol */
return 0;
}
static void handle_symbol(struct module *mod, struct elf_info *info,
const Elf_Sym *sym, const char *symname)
{
switch (sym->st_shndx) {
case SHN_COMMON:
if (strstarts(symname, "__gnu_lto_")) {
/* Should warn here, but modpost runs before the linker */
} else
warn("\"%s\" [%s] is COMMON symbol\n", symname, mod->name);
break;
case SHN_UNDEF:
/* undefined symbol */
if (ELF_ST_BIND(sym->st_info) != STB_GLOBAL &&
ELF_ST_BIND(sym->st_info) != STB_WEAK)
break;
if (ignore_undef_symbol(info, symname))
break;
if (info->hdr->e_machine == EM_SPARC ||
info->hdr->e_machine == EM_SPARCV9) {
/* Ignore register directives. */
if (ELF_ST_TYPE(sym->st_info) == STT_SPARC_REGISTER)
break;
if (symname[0] == '.') {
char *munged = NOFAIL(strdup(symname));
munged[0] = '_';
munged[1] = toupper(munged[1]);
symname = munged;
}
}
sym_add_unresolved(symname, mod,
ELF_ST_BIND(sym->st_info) == STB_WEAK);
break;
default:
if (strcmp(symname, "init_module") == 0)
mod->has_init = true;
if (strcmp(symname, "cleanup_module") == 0)
mod->has_cleanup = true;
break;
}
}
/**
* Parse tag=value strings from .modinfo section
**/
static char *next_string(char *string, unsigned long *secsize)
{
/* Skip non-zero chars */
while (string[0]) {
string++;
if ((*secsize)-- <= 1)
return NULL;
}
/* Skip any zero padding. */
while (!string[0]) {
string++;
if ((*secsize)-- <= 1)
return NULL;
}
return string;
}
static char *get_next_modinfo(struct elf_info *info, const char *tag,
char *prev)
{
char *p;
unsigned int taglen = strlen(tag);
char *modinfo = info->modinfo;
unsigned long size = info->modinfo_len;
if (prev) {
size -= prev - modinfo;
modinfo = next_string(prev, &size);
}
for (p = modinfo; p; p = next_string(p, &size)) {
if (strncmp(p, tag, taglen) == 0 && p[taglen] == '=')
return p + taglen + 1;
}
return NULL;
}
static char *get_modinfo(struct elf_info *info, const char *tag)
{
return get_next_modinfo(info, tag, NULL);
}
static const char *sym_name(struct elf_info *elf, Elf_Sym *sym)
{
if (sym)
return elf->strtab + sym->st_name;
else
return "(unknown)";
}
/*
* Check whether the 'string' argument matches one of the 'patterns',
* an array of shell wildcard patterns (glob).
*
* Return true is there is a match.
*/
static bool match(const char *string, const char *const patterns[])
{
const char *pattern;
while ((pattern = *patterns++)) {
if (!fnmatch(pattern, string, 0))
return true;
}
return false;
}
/* useful to pass patterns to match() directly */
#define PATTERNS(...) \
({ \
static const char *const patterns[] = {__VA_ARGS__, NULL}; \
patterns; \
})
/* sections that we do not want to do full section mismatch check on */
static const char *const section_white_list[] =
{
".comment*",
".debug*",
".zdebug*", /* Compressed debug sections. */
".GCC.command.line", /* record-gcc-switches */
".mdebug*", /* alpha, score, mips etc. */
".pdr", /* alpha, score, mips etc. */
".stab*",
".note*",
".got*",
".toc*",
".xt.prop", /* xtensa */
".xt.lit", /* xtensa */
".arcextmap*", /* arc */
".gnu.linkonce.arcext*", /* arc : modules */
".cmem*", /* EZchip */
".fmt_slot*", /* EZchip */
".gnu.lto*",
".discard.*",
".llvm.call-graph-profile", /* call graph */
NULL
};
/*
* This is used to find sections missing the SHF_ALLOC flag.
* The cause of this is often a section specified in assembler
* without "ax" / "aw".
*/
static void check_section(const char *modname, struct elf_info *elf,
Elf_Shdr *sechdr)
{
const char *sec = sech_name(elf, sechdr);
if (sechdr->sh_type == SHT_PROGBITS &&
!(sechdr->sh_flags & SHF_ALLOC) &&
!match(sec, section_white_list)) {
warn("%s (%s): unexpected non-allocatable section.\n"
"Did you forget to use \"ax\"/\"aw\" in a .S file?\n"
"Note that for example <linux/init.h> contains\n"
"section definitions for use in .S files.\n\n",
modname, sec);
}
}
#define ALL_INIT_DATA_SECTIONS \
".init.setup", ".init.rodata", ".meminit.rodata", \
".init.data", ".meminit.data"
#define ALL_PCI_INIT_SECTIONS \
".pci_fixup_early", ".pci_fixup_header", ".pci_fixup_final", \
".pci_fixup_enable", ".pci_fixup_resume", \
".pci_fixup_resume_early", ".pci_fixup_suspend"
#define ALL_XXXINIT_SECTIONS ".meminit.*"
#define ALL_INIT_SECTIONS INIT_SECTIONS, ALL_XXXINIT_SECTIONS
#define ALL_EXIT_SECTIONS ".exit.*"
#define DATA_SECTIONS ".data", ".data.rel"
#define TEXT_SECTIONS ".text", ".text.*", ".sched.text", \
".kprobes.text", ".cpuidle.text", ".noinstr.text", \
".ltext", ".ltext.*"
#define OTHER_TEXT_SECTIONS ".ref.text", ".head.text", ".spinlock.text", \
".fixup", ".entry.text", ".exception.text", \
powerpc/bug: Provide better flexibility to WARN_ON/__WARN_FLAGS() with asm goto Using asm goto in __WARN_FLAGS() and WARN_ON() allows more flexibility to GCC. For that add an entry to the exception table so that program_check_exception() knowns where to resume execution after a WARNING. Here are two exemples. The first one is done on PPC32 (which benefits from the previous patch), the second is on PPC64. unsigned long test(struct pt_regs *regs) { int ret; WARN_ON(regs->msr & MSR_PR); return regs->gpr[3]; } unsigned long test9w(unsigned long a, unsigned long b) { if (WARN_ON(!b)) return 0; return a / b; } Before the patch: 000003a8 <test>: 3a8: 81 23 00 84 lwz r9,132(r3) 3ac: 71 29 40 00 andi. r9,r9,16384 3b0: 40 82 00 0c bne 3bc <test+0x14> 3b4: 80 63 00 0c lwz r3,12(r3) 3b8: 4e 80 00 20 blr 3bc: 0f e0 00 00 twui r0,0 3c0: 80 63 00 0c lwz r3,12(r3) 3c4: 4e 80 00 20 blr 0000000000000bf0 <.test9w>: bf0: 7c 89 00 74 cntlzd r9,r4 bf4: 79 29 d1 82 rldicl r9,r9,58,6 bf8: 0b 09 00 00 tdnei r9,0 bfc: 2c 24 00 00 cmpdi r4,0 c00: 41 82 00 0c beq c0c <.test9w+0x1c> c04: 7c 63 23 92 divdu r3,r3,r4 c08: 4e 80 00 20 blr c0c: 38 60 00 00 li r3,0 c10: 4e 80 00 20 blr After the patch: 000003a8 <test>: 3a8: 81 23 00 84 lwz r9,132(r3) 3ac: 71 29 40 00 andi. r9,r9,16384 3b0: 40 82 00 0c bne 3bc <test+0x14> 3b4: 80 63 00 0c lwz r3,12(r3) 3b8: 4e 80 00 20 blr 3bc: 0f e0 00 00 twui r0,0 0000000000000c50 <.test9w>: c50: 7c 89 00 74 cntlzd r9,r4 c54: 79 29 d1 82 rldicl r9,r9,58,6 c58: 0b 09 00 00 tdnei r9,0 c5c: 7c 63 23 92 divdu r3,r3,r4 c60: 4e 80 00 20 blr c70: 38 60 00 00 li r3,0 c74: 4e 80 00 20 blr In the first exemple, we see GCC doesn't need to duplicate what happens after the trap. In the second exemple, we see that GCC doesn't need to emit a test and a branch in the likely path in addition to the trap. We've got some WARN_ON() in .softirqentry.text section so it needs to be added in the OTHER_TEXT_SECTIONS in modpost.c Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu> Signed-off-by: Michael Ellerman <mpe@ellerman.id.au> Link: https://lore.kernel.org/r/389962b1b702e3c78d169e59bcfac56282889173.1618331882.git.christophe.leroy@csgroup.eu
2021-04-13 16:38:10 +00:00
".coldtext", ".softirqentry.text"
#define INIT_SECTIONS ".init.*"
#define ALL_TEXT_SECTIONS ".init.text", ".meminit.text", ".exit.text", \
TEXT_SECTIONS, OTHER_TEXT_SECTIONS
kbuild: add verbose option to Section mismatch reporting in modpost If the config option CONFIG_SECTION_MISMATCH is not set and we see a Section mismatch present the following to the user: modpost: Found 1 section mismatch(es). To see additional details select "Enable full Section mismatch analysis" in the Kernel Hacking menu (CONFIG_SECTION_MISMATCH). If the option CONFIG_SECTION_MISMATCH is selected then be verbose in the Section mismatch reporting from mdopost. Sample outputs: WARNING: o-x86_64/vmlinux.o(.text+0x7396): Section mismatch in reference from the function discover_ebda() to the variable .init.data:ebda_addr The function discover_ebda() references the variable __initdata ebda_addr. This is often because discover_ebda lacks a __initdata annotation or the annotation of ebda_addr is wrong. WARNING: o-x86_64/vmlinux.o(.data+0x74d58): Section mismatch in reference from the variable pci_serial_quirks to the function .devexit.text:pci_plx9050_exit() The variable pci_serial_quirks references the function __devexit pci_plx9050_exit() If the reference is valid then annotate the variable with __exit* (see linux/init.h) or name the variable: *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, WARNING: o-x86_64/vmlinux.o(__ksymtab+0x630): Section mismatch in reference from the variable __ksymtab_arch_register_cpu to the function .cpuinit.text:arch_register_cpu() The symbol arch_register_cpu is exported and annotated __cpuinit Fix this by removing the __cpuinit annotation of arch_register_cpu or drop the export. Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
2008-01-24 20:12:37 +00:00
enum mismatch {
TEXTDATA_TO_ANY_INIT_EXIT,
XXXINIT_TO_SOME_INIT,
ANY_INIT_TO_ANY_EXIT,
ANY_EXIT_TO_ANY_INIT,
EXTABLE_TO_NON_TEXT,
kbuild: add verbose option to Section mismatch reporting in modpost If the config option CONFIG_SECTION_MISMATCH is not set and we see a Section mismatch present the following to the user: modpost: Found 1 section mismatch(es). To see additional details select "Enable full Section mismatch analysis" in the Kernel Hacking menu (CONFIG_SECTION_MISMATCH). If the option CONFIG_SECTION_MISMATCH is selected then be verbose in the Section mismatch reporting from mdopost. Sample outputs: WARNING: o-x86_64/vmlinux.o(.text+0x7396): Section mismatch in reference from the function discover_ebda() to the variable .init.data:ebda_addr The function discover_ebda() references the variable __initdata ebda_addr. This is often because discover_ebda lacks a __initdata annotation or the annotation of ebda_addr is wrong. WARNING: o-x86_64/vmlinux.o(.data+0x74d58): Section mismatch in reference from the variable pci_serial_quirks to the function .devexit.text:pci_plx9050_exit() The variable pci_serial_quirks references the function __devexit pci_plx9050_exit() If the reference is valid then annotate the variable with __exit* (see linux/init.h) or name the variable: *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, WARNING: o-x86_64/vmlinux.o(__ksymtab+0x630): Section mismatch in reference from the variable __ksymtab_arch_register_cpu to the function .cpuinit.text:arch_register_cpu() The symbol arch_register_cpu is exported and annotated __cpuinit Fix this by removing the __cpuinit annotation of arch_register_cpu or drop the export. Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
2008-01-24 20:12:37 +00:00
};
/**
* Describe how to match sections on different criteria:
*
* @fromsec: Array of sections to be matched.
*
* @bad_tosec: Relocations applied to a section in @fromsec to a section in
* this array is forbidden (black-list). Can be empty.
*
* @good_tosec: Relocations applied to a section in @fromsec must be
* targeting sections in this array (white-list). Can be empty.
*
* @mismatch: Type of mismatch.
*/
struct sectioncheck {
const char *fromsec[20];
const char *bad_tosec[20];
const char *good_tosec[20];
kbuild: add verbose option to Section mismatch reporting in modpost If the config option CONFIG_SECTION_MISMATCH is not set and we see a Section mismatch present the following to the user: modpost: Found 1 section mismatch(es). To see additional details select "Enable full Section mismatch analysis" in the Kernel Hacking menu (CONFIG_SECTION_MISMATCH). If the option CONFIG_SECTION_MISMATCH is selected then be verbose in the Section mismatch reporting from mdopost. Sample outputs: WARNING: o-x86_64/vmlinux.o(.text+0x7396): Section mismatch in reference from the function discover_ebda() to the variable .init.data:ebda_addr The function discover_ebda() references the variable __initdata ebda_addr. This is often because discover_ebda lacks a __initdata annotation or the annotation of ebda_addr is wrong. WARNING: o-x86_64/vmlinux.o(.data+0x74d58): Section mismatch in reference from the variable pci_serial_quirks to the function .devexit.text:pci_plx9050_exit() The variable pci_serial_quirks references the function __devexit pci_plx9050_exit() If the reference is valid then annotate the variable with __exit* (see linux/init.h) or name the variable: *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, WARNING: o-x86_64/vmlinux.o(__ksymtab+0x630): Section mismatch in reference from the variable __ksymtab_arch_register_cpu to the function .cpuinit.text:arch_register_cpu() The symbol arch_register_cpu is exported and annotated __cpuinit Fix this by removing the __cpuinit annotation of arch_register_cpu or drop the export. Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
2008-01-24 20:12:37 +00:00
enum mismatch mismatch;
};
static const struct sectioncheck sectioncheck[] = {
/* Do not reference init/exit code/data from
* normal code and data
*/
kbuild: add verbose option to Section mismatch reporting in modpost If the config option CONFIG_SECTION_MISMATCH is not set and we see a Section mismatch present the following to the user: modpost: Found 1 section mismatch(es). To see additional details select "Enable full Section mismatch analysis" in the Kernel Hacking menu (CONFIG_SECTION_MISMATCH). If the option CONFIG_SECTION_MISMATCH is selected then be verbose in the Section mismatch reporting from mdopost. Sample outputs: WARNING: o-x86_64/vmlinux.o(.text+0x7396): Section mismatch in reference from the function discover_ebda() to the variable .init.data:ebda_addr The function discover_ebda() references the variable __initdata ebda_addr. This is often because discover_ebda lacks a __initdata annotation or the annotation of ebda_addr is wrong. WARNING: o-x86_64/vmlinux.o(.data+0x74d58): Section mismatch in reference from the variable pci_serial_quirks to the function .devexit.text:pci_plx9050_exit() The variable pci_serial_quirks references the function __devexit pci_plx9050_exit() If the reference is valid then annotate the variable with __exit* (see linux/init.h) or name the variable: *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, WARNING: o-x86_64/vmlinux.o(__ksymtab+0x630): Section mismatch in reference from the variable __ksymtab_arch_register_cpu to the function .cpuinit.text:arch_register_cpu() The symbol arch_register_cpu is exported and annotated __cpuinit Fix this by removing the __cpuinit annotation of arch_register_cpu or drop the export. Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
2008-01-24 20:12:37 +00:00
{
.fromsec = { TEXT_SECTIONS, DATA_SECTIONS, NULL },
.bad_tosec = { ALL_INIT_SECTIONS, ALL_EXIT_SECTIONS, NULL },
.mismatch = TEXTDATA_TO_ANY_INIT_EXIT,
},
/* Do not reference init code/data from meminit code/data */
{
.fromsec = { ALL_XXXINIT_SECTIONS, NULL },
.bad_tosec = { INIT_SECTIONS, NULL },
.mismatch = XXXINIT_TO_SOME_INIT,
},
/* Do not use exit code/data from init code */
{
.fromsec = { ALL_INIT_SECTIONS, NULL },
.bad_tosec = { ALL_EXIT_SECTIONS, NULL },
.mismatch = ANY_INIT_TO_ANY_EXIT,
},
/* Do not use init code/data from exit code */
{
.fromsec = { ALL_EXIT_SECTIONS, NULL },
.bad_tosec = { ALL_INIT_SECTIONS, NULL },
.mismatch = ANY_EXIT_TO_ANY_INIT,
},
{
.fromsec = { ALL_PCI_INIT_SECTIONS, NULL },
.bad_tosec = { INIT_SECTIONS, NULL },
.mismatch = ANY_INIT_TO_ANY_EXIT,
},
{
.fromsec = { "__ex_table", NULL },
/* If you're adding any new black-listed sections in here, consider
* adding a special 'printer' for them in scripts/check_extable.
*/
.bad_tosec = { ".altinstr_replacement", NULL },
.good_tosec = {ALL_TEXT_SECTIONS , NULL},
.mismatch = EXTABLE_TO_NON_TEXT,
}
};
static const struct sectioncheck *section_mismatch(
const char *fromsec, const char *tosec)
{
int i;
/*
* The target section could be the SHT_NUL section when we're
* handling relocations to un-resolved symbols, trying to match it
* doesn't make much sense and causes build failures on parisc
* architectures.
*/
if (*tosec == '\0')
return NULL;
for (i = 0; i < ARRAY_SIZE(sectioncheck); i++) {
const struct sectioncheck *check = &sectioncheck[i];
if (match(fromsec, check->fromsec)) {
if (check->bad_tosec[0] && match(tosec, check->bad_tosec))
return check;
if (check->good_tosec[0] && !match(tosec, check->good_tosec))
return check;
}
}
return NULL;
}
/**
* Whitelist to allow certain references to pass with no warning.
*
* Pattern 1:
* If a module parameter is declared __initdata and permissions=0
* then this is legal despite the warning generated.
* We cannot see value of permissions here, so just ignore
* this pattern.
* The pattern is identified by:
* tosec = .init.data
* fromsec = .data*
* atsym =__param*
*
* Pattern 1a:
* module_param_call() ops can refer to __init set function if permissions=0
* The pattern is identified by:
* tosec = .init.text
* fromsec = .data*
* atsym = __param_ops_*
*
[PATCH] x86-64: Modpost whitelist reference to more symbols (pattern 3) o MODPOST generates warning on i386 if kernel is compiled with CONFIG_RELOCATABLE=y. WARNING: vmlinux - Section mismatch: reference to .init.text:__init_begin from .text between 'free_initmem' (at offset 0xc0114fd3) and 'do_test_wp_bit' WARNING: vmlinux - Section mismatch: reference to .init.text:_sinittext from .text between 'core_kernel_text' (at offset 0xc012aeae) and 'kernel_text_address' WARNING: vmlinux - Section mismatch: reference to .init.text:_einittext from .text between 'core_kernel_text' (at offset 0xc012aeb7) and 'kernel_text_address' WARNING: vmlinux - Section mismatch: reference to .init.text:_sinittext from .text between 'get_symbol_pos' (at offset 0xc0135776) and 'reset_iter' WARNING: vmlinux - Section mismatch: reference to .init.text:_einittext from .text between 'get_symbol_pos' (at offset 0xc013577d) and 'reset_iter' o These symbols (__init_begin, _sinittext, _einittext) belong to init section and generally represent a section boundary. These are special symbols in the sense that their size is zero and no memory is allocated for them in init section. Their addr and value are same. So even if we free the init section, it is ok to reference them. o Whitelist access to such select symbols in MODPOST. Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Andi Kleen <ak@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org>
2007-01-11 00:52:44 +00:00
* Pattern 3:
* Whitelist all references from .head.text to any init section
*
* Pattern 4:
[PATCH] x86-64: Modpost whitelist reference to more symbols (pattern 3) o MODPOST generates warning on i386 if kernel is compiled with CONFIG_RELOCATABLE=y. WARNING: vmlinux - Section mismatch: reference to .init.text:__init_begin from .text between 'free_initmem' (at offset 0xc0114fd3) and 'do_test_wp_bit' WARNING: vmlinux - Section mismatch: reference to .init.text:_sinittext from .text between 'core_kernel_text' (at offset 0xc012aeae) and 'kernel_text_address' WARNING: vmlinux - Section mismatch: reference to .init.text:_einittext from .text between 'core_kernel_text' (at offset 0xc012aeb7) and 'kernel_text_address' WARNING: vmlinux - Section mismatch: reference to .init.text:_sinittext from .text between 'get_symbol_pos' (at offset 0xc0135776) and 'reset_iter' WARNING: vmlinux - Section mismatch: reference to .init.text:_einittext from .text between 'get_symbol_pos' (at offset 0xc013577d) and 'reset_iter' o These symbols (__init_begin, _sinittext, _einittext) belong to init section and generally represent a section boundary. These are special symbols in the sense that their size is zero and no memory is allocated for them in init section. Their addr and value are same. So even if we free the init section, it is ok to reference them. o Whitelist access to such select symbols in MODPOST. Signed-off-by: Vivek Goyal <vgoyal@in.ibm.com> Signed-off-by: Andi Kleen <ak@suse.de> Cc: "Eric W. Biederman" <ebiederm@xmission.com> Cc: Andi Kleen <ak@suse.de> Signed-off-by: Andrew Morton <akpm@osdl.org>
2007-01-11 00:52:44 +00:00
* Some symbols belong to init section but still it is ok to reference
* these from non-init sections as these symbols don't have any memory
* allocated for them and symbol address and value are same. So even
* if init section is freed, its ok to reference those symbols.
* For ex. symbols marking the init section boundaries.
* This pattern is identified by
* refsymname = __init_begin, _sinittext, _einittext
*
* Pattern 5:
* GCC may optimize static inlines when fed constant arg(s) resulting
* in functions like cpumask_empty() -- generating an associated symbol
* cpumask_empty.constprop.3 that appears in the audit. If the const that
* is passed in comes from __init, like say nmi_ipi_mask, we get a
* meaningless section warning. May need to add isra symbols too...
* This pattern is identified by
* tosec = init section
* fromsec = text section
* refsymname = *.constprop.*
*
**/
static int secref_whitelist(const char *fromsec, const char *fromsym,
const char *tosec, const char *tosym)
{
/* Check for pattern 1 */
if (match(tosec, PATTERNS(ALL_INIT_DATA_SECTIONS)) &&
match(fromsec, PATTERNS(DATA_SECTIONS)) &&
strstarts(fromsym, "__param"))
return 0;
/* Check for pattern 1a */
if (strcmp(tosec, ".init.text") == 0 &&
match(fromsec, PATTERNS(DATA_SECTIONS)) &&
strstarts(fromsym, "__param_ops_"))
return 0;
/* symbols in data sections that may refer to any init/exit sections */
if (match(fromsec, PATTERNS(DATA_SECTIONS)) &&
match(tosec, PATTERNS(ALL_INIT_SECTIONS, ALL_EXIT_SECTIONS)) &&
match(fromsym, PATTERNS("*_ops", "*_probe", "*_console")))
return 0;
/*
* symbols in data sections must not refer to .exit.*, but there are
* quite a few offenders, so hide these unless for W=1 builds until
* these are fixed.
*/
if (!extra_warn &&
match(fromsec, PATTERNS(DATA_SECTIONS)) &&
match(tosec, PATTERNS(ALL_EXIT_SECTIONS)) &&
match(fromsym, PATTERNS("*driver")))
return 0;
/* Check for pattern 3 */
if (strstarts(fromsec, ".head.text") &&
match(tosec, PATTERNS(ALL_INIT_SECTIONS)))
return 0;
/* Check for pattern 4 */
if (match(tosym, PATTERNS("__init_begin", "_sinittext", "_einittext")))
return 0;
/* Check for pattern 5 */
if (match(fromsec, PATTERNS(ALL_TEXT_SECTIONS)) &&
match(tosec, PATTERNS(ALL_INIT_SECTIONS)) &&
match(fromsym, PATTERNS("*.constprop.*")))
modpost: skip ELF local symbols during section mismatch check During development of a serial console driver with a gcc 8.2.0 toolchain for RISC-V, the following modpost warning appeared: ---- WARNING: vmlinux.o(.data+0x19b10): Section mismatch in reference from the variable .LANCHOR1 to the function .init.text:sifive_serial_console_setup() The variable .LANCHOR1 references the function __init sifive_serial_console_setup() If the reference is valid then annotate the variable with __init* or __refdata (see linux/init.h) or name the variable: *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console ---- ".LANCHOR1" is an ELF local symbol, automatically created by gcc's section anchor generation code: https://gcc.gnu.org/onlinedocs/gccint/Anchored-Addresses.html https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=gcc/varasm.c;h=cd9591a45617464946dcf9a126dde277d9de9804;hb=9fb89fa845c1b2e0a18d85ada0b077c84508ab78#l7473 This was verified by compiling the kernel with -fno-section-anchors and observing that the ".LANCHOR1" ELF local symbol disappeared, and modpost no longer warned about the section mismatch. The serial driver code idiom triggering the warning is standard Linux serial driver practice that has a specific whitelist inclusion in modpost.c. I'm neither a modpost nor an ELF expert, but naively, it doesn't seem useful for modpost to report section mismatch warnings caused by ELF local symbols by default. Local symbols have compiler-generated names, and thus bypass modpost's whitelisting algorithm, which relies on the presence of a non-autogenerated symbol name. This increases the likelihood that false positive warnings will be generated (as in the above case). Thus, disable section mismatch reporting on ELF local symbols. The rationale here is similar to that of commit 2e3a10a1551d ("ARM: avoid ARM binutils leaking ELF local symbols") and of similar code already present in modpost.c: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/scripts/mod/modpost.c?h=v4.19-rc4&id=7876320f88802b22d4e2daf7eb027dd14175a0f8#n1256 This third version of the patch implements a suggestion from Masahiro Yamada <yamada.masahiro@socionext.com> to restructure the code as an additional pattern matching step inside secref_whitelist(), and further improves the patch description. Signed-off-by: Paul Walmsley <paul.walmsley@sifive.com> Signed-off-by: Paul Walmsley <paul@pwsan.com> Acked-by: Sam Ravnborg <sam@ravnborg.org> Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
2018-11-21 21:14:13 +00:00
return 0;
return 1;
}
static Elf_Sym *find_fromsym(struct elf_info *elf, Elf_Addr addr,
unsigned int secndx)
kbuild: add verbose option to Section mismatch reporting in modpost If the config option CONFIG_SECTION_MISMATCH is not set and we see a Section mismatch present the following to the user: modpost: Found 1 section mismatch(es). To see additional details select "Enable full Section mismatch analysis" in the Kernel Hacking menu (CONFIG_SECTION_MISMATCH). If the option CONFIG_SECTION_MISMATCH is selected then be verbose in the Section mismatch reporting from mdopost. Sample outputs: WARNING: o-x86_64/vmlinux.o(.text+0x7396): Section mismatch in reference from the function discover_ebda() to the variable .init.data:ebda_addr The function discover_ebda() references the variable __initdata ebda_addr. This is often because discover_ebda lacks a __initdata annotation or the annotation of ebda_addr is wrong. WARNING: o-x86_64/vmlinux.o(.data+0x74d58): Section mismatch in reference from the variable pci_serial_quirks to the function .devexit.text:pci_plx9050_exit() The variable pci_serial_quirks references the function __devexit pci_plx9050_exit() If the reference is valid then annotate the variable with __exit* (see linux/init.h) or name the variable: *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, WARNING: o-x86_64/vmlinux.o(__ksymtab+0x630): Section mismatch in reference from the variable __ksymtab_arch_register_cpu to the function .cpuinit.text:arch_register_cpu() The symbol arch_register_cpu is exported and annotated __cpuinit Fix this by removing the __cpuinit annotation of arch_register_cpu or drop the export. Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
2008-01-24 20:12:37 +00:00
{
return symsearch_find_nearest(elf, addr, secndx, false, ~0);
kbuild: add verbose option to Section mismatch reporting in modpost If the config option CONFIG_SECTION_MISMATCH is not set and we see a Section mismatch present the following to the user: modpost: Found 1 section mismatch(es). To see additional details select "Enable full Section mismatch analysis" in the Kernel Hacking menu (CONFIG_SECTION_MISMATCH). If the option CONFIG_SECTION_MISMATCH is selected then be verbose in the Section mismatch reporting from mdopost. Sample outputs: WARNING: o-x86_64/vmlinux.o(.text+0x7396): Section mismatch in reference from the function discover_ebda() to the variable .init.data:ebda_addr The function discover_ebda() references the variable __initdata ebda_addr. This is often because discover_ebda lacks a __initdata annotation or the annotation of ebda_addr is wrong. WARNING: o-x86_64/vmlinux.o(.data+0x74d58): Section mismatch in reference from the variable pci_serial_quirks to the function .devexit.text:pci_plx9050_exit() The variable pci_serial_quirks references the function __devexit pci_plx9050_exit() If the reference is valid then annotate the variable with __exit* (see linux/init.h) or name the variable: *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, WARNING: o-x86_64/vmlinux.o(__ksymtab+0x630): Section mismatch in reference from the variable __ksymtab_arch_register_cpu to the function .cpuinit.text:arch_register_cpu() The symbol arch_register_cpu is exported and annotated __cpuinit Fix this by removing the __cpuinit annotation of arch_register_cpu or drop the export. Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
2008-01-24 20:12:37 +00:00
}
static Elf_Sym *find_tosym(struct elf_info *elf, Elf_Addr addr, Elf_Sym *sym)
{
Elf_Sym *new_sym;
/* If the supplied symbol has a valid name, return it */
if (is_valid_name(elf, sym))
return sym;
/*
* Strive to find a better symbol name, but the resulting name may not
* match the symbol referenced in the original code.
*/
new_sym = symsearch_find_nearest(elf, addr, get_secindex(elf, sym),
true, 20);
return new_sym ? new_sym : sym;
}
static bool is_executable_section(struct elf_info *elf, unsigned int secndx)
kbuild: add verbose option to Section mismatch reporting in modpost If the config option CONFIG_SECTION_MISMATCH is not set and we see a Section mismatch present the following to the user: modpost: Found 1 section mismatch(es). To see additional details select "Enable full Section mismatch analysis" in the Kernel Hacking menu (CONFIG_SECTION_MISMATCH). If the option CONFIG_SECTION_MISMATCH is selected then be verbose in the Section mismatch reporting from mdopost. Sample outputs: WARNING: o-x86_64/vmlinux.o(.text+0x7396): Section mismatch in reference from the function discover_ebda() to the variable .init.data:ebda_addr The function discover_ebda() references the variable __initdata ebda_addr. This is often because discover_ebda lacks a __initdata annotation or the annotation of ebda_addr is wrong. WARNING: o-x86_64/vmlinux.o(.data+0x74d58): Section mismatch in reference from the variable pci_serial_quirks to the function .devexit.text:pci_plx9050_exit() The variable pci_serial_quirks references the function __devexit pci_plx9050_exit() If the reference is valid then annotate the variable with __exit* (see linux/init.h) or name the variable: *driver, *_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console, WARNING: o-x86_64/vmlinux.o(__ksymtab+0x630): Section mismatch in reference from the variable __ksymtab_arch_register_cpu to the function .cpuinit.text:arch_register_cpu() The symbol arch_register_cpu is exported and annotated __cpuinit Fix this by removing the __cpuinit annotation of arch_register_cpu or drop the export. Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
2008-01-24 20:12:37 +00:00
{
if (secndx >= elf->num_sections)
return false;
return (elf->sechdrs[secndx].sh_flags & SHF_EXECINSTR) != 0;
}
static void default_mismatch_handler(const char *modname, struct elf_info *elf,
const struct sectioncheck* const mismatch,
Elf_Sym *tsym,
unsigned int fsecndx, const char *fromsec, Elf_Addr faddr,
const char *tosec, Elf_Addr taddr)
{
Elf_Sym *from;
const char *tosym;
const char *fromsym;
from = find_fromsym(elf, faddr, fsecndx);
fromsym = sym_name(elf, from);
tsym = find_tosym(elf, taddr, tsym);
tosym = sym_name(elf, tsym);
/* check whitelist - we may ignore it */
if (!secref_whitelist(fromsec, fromsym, tosec, tosym))
return;
sec_mismatch_count++;
warn("%s: section mismatch in reference: %s+0x%x (section: %s) -> %s (section: %s)\n",
modname, fromsym,
(unsigned int)(faddr - (from ? from->st_value : 0)),
fromsec, tosym, tosec);
if (mismatch->mismatch == EXTABLE_TO_NON_TEXT) {
if (match(tosec, mismatch->bad_tosec))
fatal("The relocation at %s+0x%lx references\n"
"section \"%s\" which is black-listed.\n"
"Something is seriously wrong and should be fixed.\n"
"You might get more information about where this is\n"
"coming from by using scripts/check_extable.sh %s\n",
fromsec, (long)faddr, tosec, modname);
else if (is_executable_section(elf, get_secindex(elf, tsym)))
warn("The relocation at %s+0x%lx references\n"
"section \"%s\" which is not in the list of\n"
"authorized sections. If you're adding a new section\n"
"and/or if this reference is valid, add \"%s\" to the\n"
"list of authorized sections to jump to on fault.\n"
"This can be achieved by adding \"%s\" to\n"
"OTHER_TEXT_SECTIONS in scripts/mod/modpost.c.\n",
fromsec, (long)faddr, tosec, tosec, tosec);
else
error("%s+0x%lx references non-executable section '%s'\n",
fromsec, (long)faddr, tosec);
}
}
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
static void check_export_symbol(struct module *mod, struct elf_info *elf,
Elf_Addr faddr, const char *secname,
Elf_Sym *sym)
{
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
static const char *prefix = "__export_symbol_";
const char *label_name, *name, *data;
Elf_Sym *label;
struct symbol *s;
bool is_gpl;
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
label = find_fromsym(elf, faddr, elf->export_symbol_secndx);
label_name = sym_name(elf, label);
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
if (!strstarts(label_name, prefix)) {
error("%s: .export_symbol section contains strange symbol '%s'\n",
mod->name, label_name);
return;
}
if (ELF_ST_BIND(sym->st_info) != STB_GLOBAL &&
ELF_ST_BIND(sym->st_info) != STB_WEAK) {
error("%s: local symbol '%s' was exported\n", mod->name,
label_name + strlen(prefix));
return;
}
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
name = sym_name(elf, sym);
if (strcmp(label_name + strlen(prefix), name)) {
error("%s: .export_symbol section references '%s', but it does not seem to be an export symbol\n",
mod->name, name);
return;
}
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
data = sym_get_data(elf, label); /* license */
if (!strcmp(data, "GPL")) {
is_gpl = true;
} else if (!strcmp(data, "")) {
is_gpl = false;
} else {
error("%s: unknown license '%s' was specified for '%s'\n",
mod->name, data, name);
return;
}
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
data += strlen(data) + 1; /* namespace */
s = sym_add_exported(name, mod, is_gpl, data);
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
/*
* We need to be aware whether we are exporting a function or
* a data on some architectures.
*/
s->is_func = (ELF_ST_TYPE(sym->st_info) == STT_FUNC);
/*
* For parisc64, symbols prefixed $$ from the library have the symbol type
* STT_LOPROC. They should be handled as functions too.
*/
if (elf->hdr->e_ident[EI_CLASS] == ELFCLASS64 &&
elf->hdr->e_machine == EM_PARISC &&
ELF_ST_TYPE(sym->st_info) == STT_LOPROC)
s->is_func = true;
if (match(secname, PATTERNS(ALL_INIT_SECTIONS)))
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
warn("%s: %s: EXPORT_SYMBOL used for init symbol. Remove __init or EXPORT_SYMBOL.\n",
mod->name, name);
else if (match(secname, PATTERNS(ALL_EXIT_SECTIONS)))
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
warn("%s: %s: EXPORT_SYMBOL used for exit symbol. Remove __exit or EXPORT_SYMBOL.\n",
mod->name, name);
}
static void check_section_mismatch(struct module *mod, struct elf_info *elf,
Elf_Sym *sym,
unsigned int fsecndx, const char *fromsec,
Elf_Addr faddr, Elf_Addr taddr)
{
const char *tosec = sec_name(elf, get_secindex(elf, sym));
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
const struct sectioncheck *mismatch;
if (module_enabled && elf->export_symbol_secndx == fsecndx) {
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
check_export_symbol(mod, elf, faddr, tosec, sym);
return;
}
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
mismatch = section_mismatch(fromsec, tosec);
if (!mismatch)
return;
default_mismatch_handler(mod->name, elf, mismatch, sym,
fsecndx, fromsec, faddr,
tosec, taddr);
}
static Elf_Addr addend_386_rel(uint32_t *location, unsigned int r_type)
{
switch (r_type) {
case R_386_32:
return TO_NATIVE(*location);
case R_386_PC32:
return TO_NATIVE(*location) + 4;
}
return (Elf_Addr)(-1);
}
#ifndef R_ARM_CALL
#define R_ARM_CALL 28
#endif
#ifndef R_ARM_JUMP24
#define R_ARM_JUMP24 29
#endif
#ifndef R_ARM_THM_CALL
#define R_ARM_THM_CALL 10
#endif
#ifndef R_ARM_THM_JUMP24
#define R_ARM_THM_JUMP24 30
#endif
#ifndef R_ARM_MOVW_ABS_NC
#define R_ARM_MOVW_ABS_NC 43
#endif
#ifndef R_ARM_MOVT_ABS
#define R_ARM_MOVT_ABS 44
#endif
#ifndef R_ARM_THM_MOVW_ABS_NC
#define R_ARM_THM_MOVW_ABS_NC 47
#endif
#ifndef R_ARM_THM_MOVT_ABS
#define R_ARM_THM_MOVT_ABS 48
#endif
#ifndef R_ARM_THM_JUMP19
#define R_ARM_THM_JUMP19 51
#endif
static int32_t sign_extend32(int32_t value, int index)
{
uint8_t shift = 31 - index;
return (int32_t)(value << shift) >> shift;
}
static Elf_Addr addend_arm_rel(void *loc, Elf_Sym *sym, unsigned int r_type)
{
modpost: fix section_mismatch message for R_ARM_THM_{CALL,JUMP24,JUMP19} addend_arm_rel() processes R_ARM_THM_CALL, R_ARM_THM_JUMP24, R_ARM_THM_JUMP19 in a wrong way. Here, test code. [test code for R_ARM_THM_JUMP24]   .section .init.text,"ax"   bar:           bx      lr   .section .text,"ax"   .globl foo   foo:           b       bar [test code for R_ARM_THM_CALL]   .section .init.text,"ax"   bar:           bx      lr   .section .text,"ax"   .globl foo   foo:           push    {lr}           bl      bar           pop     {pc} If you compile it with CONFIG_THUMB2_KERNEL=y, modpost will show the symbol name, (unknown).   WARNING: modpost: vmlinux.o: section mismatch in reference: foo (section: .text) -> (unknown) (section: .init.text) (You need to use GNU linker instead of LLD to reproduce it.) Fix the code to make modpost show the correct symbol name. I checked arch/arm/kernel/module.c to learn the encoding of R_ARM_THM_CALL and R_ARM_THM_JUMP24. The module does not support R_ARM_THM_JUMP19, but I checked its encoding in ARM ARM. The '+4' is the compensation for pc-relative instruction. It is documented in "ELF for the Arm Architecture" [1].   "If the relocation is pc-relative then compensation for the PC bias   (the PC value is 8 bytes ahead of the executing instruction in Arm   state and 4 bytes in Thumb state) must be encoded in the relocation   by the object producer." [1]: https://github.com/ARM-software/abi-aa/blob/main/aaelf32/aaelf32.rst Fixes: c9698e5cd6ad ("ARM: 7964/1: Detect section mismatches in thumb relocations") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-01 12:10:00 +00:00
uint32_t inst, upper, lower, sign, j1, j2;
int32_t offset;
switch (r_type) {
case R_ARM_ABS32:
case R_ARM_REL32:
modpost: fix section mismatch message for R_ARM_ABS32 addend_arm_rel() processes R_ARM_ABS32 in a wrong way. Here, test code. [test code 1] #include <linux/init.h> int __initdata foo; int get_foo(void) { return foo; } If you compile it with ARM versatile_defconfig, modpost will show the symbol name, (unknown). WARNING: modpost: vmlinux.o: section mismatch in reference: get_foo (section: .text) -> (unknown) (section: .init.data) (You need to use GNU linker instead of LLD to reproduce it.) If you compile it for other architectures, modpost will show the correct symbol name. WARNING: modpost: vmlinux.o: section mismatch in reference: get_foo (section: .text) -> foo (section: .init.data) For R_ARM_ABS32, addend_arm_rel() sets r->r_addend to a wrong value. I just mimicked the code in arch/arm/kernel/module.c. However, there is more difficulty for ARM. Here, test code. [test code 2] #include <linux/init.h> int __initdata foo; int get_foo(void) { return foo; } int __initdata bar; int get_bar(void) { return bar; } With this commit applied, modpost will show the following messages for ARM versatile_defconfig: WARNING: modpost: vmlinux.o: section mismatch in reference: get_foo (section: .text) -> foo (section: .init.data) WARNING: modpost: vmlinux.o: section mismatch in reference: get_bar (section: .text) -> foo (section: .init.data) The reference from 'get_bar' to 'foo' seems wrong. I have no solution for this because it is true in assembly level. In the following output, relocation at 0x1c is no longer associated with 'bar'. The two relocation entries point to the same symbol, and the offset to 'bar' is encoded in the instruction 'r0, [r3, #4]'. Disassembly of section .text: 00000000 <get_foo>: 0: e59f3004 ldr r3, [pc, #4] @ c <get_foo+0xc> 4: e5930000 ldr r0, [r3] 8: e12fff1e bx lr c: 00000000 .word 0x00000000 00000010 <get_bar>: 10: e59f3004 ldr r3, [pc, #4] @ 1c <get_bar+0xc> 14: e5930004 ldr r0, [r3, #4] 18: e12fff1e bx lr 1c: 00000000 .word 0x00000000 Relocation section '.rel.text' at offset 0x244 contains 2 entries: Offset Info Type Sym.Value Sym. Name 0000000c 00000c02 R_ARM_ABS32 00000000 .init.data 0000001c 00000c02 R_ARM_ABS32 00000000 .init.data When find_elf_symbol() gets into a situation where relsym->st_name is zero, there is no guarantee to get the symbol name as written in C. I am keeping the current logic because it is useful in many architectures, but the symbol name is not always correct depending on the optimization. I left some comments in find_tosym(). Fixes: 56a974fa2d59 ("kbuild: make better section mismatch reports on arm") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-01 12:09:55 +00:00
inst = TO_NATIVE(*(uint32_t *)loc);
return inst + sym->st_value;
case R_ARM_MOVW_ABS_NC:
case R_ARM_MOVT_ABS:
inst = TO_NATIVE(*(uint32_t *)loc);
offset = sign_extend32(((inst & 0xf0000) >> 4) | (inst & 0xfff),
15);
return offset + sym->st_value;
case R_ARM_PC24:
case R_ARM_CALL:
case R_ARM_JUMP24:
inst = TO_NATIVE(*(uint32_t *)loc);
offset = sign_extend32((inst & 0x00ffffff) << 2, 25);
return offset + sym->st_value + 8;
modpost: detect section mismatch for R_ARM_THM_{MOVW_ABS_NC,MOVT_ABS} When CONFIG_THUMB2_KERNEL is enabled, modpost fails to detect some types of section mismatches. [test code] #include <linux/init.h> int __initdata foo; int get_foo(void) { return foo; } It is apparently a bad reference, but modpost does not report anything. The test code above produces the following relocations. Relocation section '.rel.text' at offset 0x1e8 contains 2 entries: Offset Info Type Sym.Value Sym. Name 00000000 0000052f R_ARM_THM_MOVW_AB 00000000 .LANCHOR0 00000004 00000530 R_ARM_THM_MOVT_AB 00000000 .LANCHOR0 Currently, R_ARM_THM_MOVW_ABS_NC and R_ARM_THM_MOVT_ABS are just skipped. Add code to handle them. I checked arch/arm/kernel/module.c to learn how the offset is encoded in the instruction. One more thing to note for Thumb instructions - the st_value is an odd value, so you need to mask the bit 0 to get the offset. Otherwise, you will get an off-by-one error in the nearest symbol look-up. It is documented in "ELF for the ARM Architecture" [1]: In addition to the normal rules for symbol values the following rules shall also apply to symbols of type STT_FUNC: * If the symbol addresses an Arm instruction, its value is the address of the instruction (in a relocatable object, the offset of the instruction from the start of the section containing it). * If the symbol addresses a Thumb instruction, its value is the address of the instruction with bit zero set (in a relocatable object, the section offset with bit zero set). * For the purposes of relocation the value used shall be the address of the instruction (st_value & ~1). [1]: https://github.com/ARM-software/abi-aa/blob/main/aaelf32/aaelf32.rst Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-01 12:09:59 +00:00
case R_ARM_THM_MOVW_ABS_NC:
case R_ARM_THM_MOVT_ABS:
upper = TO_NATIVE(*(uint16_t *)loc);
lower = TO_NATIVE(*((uint16_t *)loc + 1));
offset = sign_extend32(((upper & 0x000f) << 12) |
((upper & 0x0400) << 1) |
((lower & 0x7000) >> 4) |
(lower & 0x00ff),
15);
return offset + sym->st_value;
modpost: fix section_mismatch message for R_ARM_THM_{CALL,JUMP24,JUMP19} addend_arm_rel() processes R_ARM_THM_CALL, R_ARM_THM_JUMP24, R_ARM_THM_JUMP19 in a wrong way. Here, test code. [test code for R_ARM_THM_JUMP24]   .section .init.text,"ax"   bar:           bx      lr   .section .text,"ax"   .globl foo   foo:           b       bar [test code for R_ARM_THM_CALL]   .section .init.text,"ax"   bar:           bx      lr   .section .text,"ax"   .globl foo   foo:           push    {lr}           bl      bar           pop     {pc} If you compile it with CONFIG_THUMB2_KERNEL=y, modpost will show the symbol name, (unknown).   WARNING: modpost: vmlinux.o: section mismatch in reference: foo (section: .text) -> (unknown) (section: .init.text) (You need to use GNU linker instead of LLD to reproduce it.) Fix the code to make modpost show the correct symbol name. I checked arch/arm/kernel/module.c to learn the encoding of R_ARM_THM_CALL and R_ARM_THM_JUMP24. The module does not support R_ARM_THM_JUMP19, but I checked its encoding in ARM ARM. The '+4' is the compensation for pc-relative instruction. It is documented in "ELF for the Arm Architecture" [1].   "If the relocation is pc-relative then compensation for the PC bias   (the PC value is 8 bytes ahead of the executing instruction in Arm   state and 4 bytes in Thumb state) must be encoded in the relocation   by the object producer." [1]: https://github.com/ARM-software/abi-aa/blob/main/aaelf32/aaelf32.rst Fixes: c9698e5cd6ad ("ARM: 7964/1: Detect section mismatches in thumb relocations") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-01 12:10:00 +00:00
case R_ARM_THM_JUMP19:
/*
* Encoding T3:
* S = upper[10]
* imm6 = upper[5:0]
* J1 = lower[13]
* J2 = lower[11]
* imm11 = lower[10:0]
* imm32 = SignExtend(S:J2:J1:imm6:imm11:'0')
*/
upper = TO_NATIVE(*(uint16_t *)loc);
lower = TO_NATIVE(*((uint16_t *)loc + 1));
sign = (upper >> 10) & 1;
j1 = (lower >> 13) & 1;
j2 = (lower >> 11) & 1;
offset = sign_extend32((sign << 20) | (j2 << 19) | (j1 << 18) |
((upper & 0x03f) << 12) |
((lower & 0x07ff) << 1),
20);
return offset + sym->st_value + 4;
case R_ARM_THM_CALL:
case R_ARM_THM_JUMP24:
modpost: fix section_mismatch message for R_ARM_THM_{CALL,JUMP24,JUMP19} addend_arm_rel() processes R_ARM_THM_CALL, R_ARM_THM_JUMP24, R_ARM_THM_JUMP19 in a wrong way. Here, test code. [test code for R_ARM_THM_JUMP24]   .section .init.text,"ax"   bar:           bx      lr   .section .text,"ax"   .globl foo   foo:           b       bar [test code for R_ARM_THM_CALL]   .section .init.text,"ax"   bar:           bx      lr   .section .text,"ax"   .globl foo   foo:           push    {lr}           bl      bar           pop     {pc} If you compile it with CONFIG_THUMB2_KERNEL=y, modpost will show the symbol name, (unknown).   WARNING: modpost: vmlinux.o: section mismatch in reference: foo (section: .text) -> (unknown) (section: .init.text) (You need to use GNU linker instead of LLD to reproduce it.) Fix the code to make modpost show the correct symbol name. I checked arch/arm/kernel/module.c to learn the encoding of R_ARM_THM_CALL and R_ARM_THM_JUMP24. The module does not support R_ARM_THM_JUMP19, but I checked its encoding in ARM ARM. The '+4' is the compensation for pc-relative instruction. It is documented in "ELF for the Arm Architecture" [1].   "If the relocation is pc-relative then compensation for the PC bias   (the PC value is 8 bytes ahead of the executing instruction in Arm   state and 4 bytes in Thumb state) must be encoded in the relocation   by the object producer." [1]: https://github.com/ARM-software/abi-aa/blob/main/aaelf32/aaelf32.rst Fixes: c9698e5cd6ad ("ARM: 7964/1: Detect section mismatches in thumb relocations") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-01 12:10:00 +00:00
/*
* Encoding T4:
* S = upper[10]
* imm10 = upper[9:0]
* J1 = lower[13]
* J2 = lower[11]
* imm11 = lower[10:0]
* I1 = NOT(J1 XOR S)
* I2 = NOT(J2 XOR S)
* imm32 = SignExtend(S:I1:I2:imm10:imm11:'0')
*/
upper = TO_NATIVE(*(uint16_t *)loc);
lower = TO_NATIVE(*((uint16_t *)loc + 1));
sign = (upper >> 10) & 1;
j1 = (lower >> 13) & 1;
j2 = (lower >> 11) & 1;
offset = sign_extend32((sign << 24) |
((~(j1 ^ sign) & 1) << 23) |
((~(j2 ^ sign) & 1) << 22) |
((upper & 0x03ff) << 12) |
((lower & 0x07ff) << 1),
24);
return offset + sym->st_value + 4;
}
return (Elf_Addr)(-1);
}
static Elf_Addr addend_mips_rel(uint32_t *location, unsigned int r_type)
{
uint32_t inst;
inst = TO_NATIVE(*location);
switch (r_type) {
case R_MIPS_LO16:
return inst & 0xffff;
case R_MIPS_26:
return (inst & 0x03ffffff) << 2;
case R_MIPS_32:
return inst;
}
return (Elf_Addr)(-1);
}
#ifndef EM_RISCV
#define EM_RISCV 243
#endif
#ifndef R_RISCV_SUB32
#define R_RISCV_SUB32 39
#endif
#ifndef EM_LOONGARCH
#define EM_LOONGARCH 258
#endif
#ifndef R_LARCH_SUB32
#define R_LARCH_SUB32 55
#endif
#ifndef R_LARCH_RELAX
#define R_LARCH_RELAX 100
#endif
#ifndef R_LARCH_ALIGN
#define R_LARCH_ALIGN 102
#endif
static void get_rel_type_and_sym(struct elf_info *elf, uint64_t r_info,
unsigned int *r_type, unsigned int *r_sym)
{
typedef struct {
Elf64_Word r_sym; /* Symbol index */
unsigned char r_ssym; /* Special symbol for 2nd relocation */
unsigned char r_type3; /* 3rd relocation type */
unsigned char r_type2; /* 2nd relocation type */
unsigned char r_type; /* 1st relocation type */
} Elf64_Mips_R_Info;
bool is_64bit = (elf->hdr->e_ident[EI_CLASS] == ELFCLASS64);
if (elf->hdr->e_machine == EM_MIPS && is_64bit) {
Elf64_Mips_R_Info *mips64_r_info = (void *)&r_info;
*r_type = mips64_r_info->r_type;
*r_sym = TO_NATIVE(mips64_r_info->r_sym);
return;
}
if (is_64bit)
r_info = TO_NATIVE((Elf64_Xword)r_info);
else
r_info = TO_NATIVE((Elf32_Word)r_info);
*r_type = ELF_R_TYPE(r_info);
*r_sym = ELF_R_SYM(r_info);
}
static void section_rela(struct module *mod, struct elf_info *elf,
unsigned int fsecndx, const char *fromsec,
const Elf_Rela *start, const Elf_Rela *stop)
{
const Elf_Rela *rela;
for (rela = start; rela < stop; rela++) {
Elf_Sym *tsym;
Elf_Addr taddr, r_offset;
unsigned int r_type, r_sym;
r_offset = TO_NATIVE(rela->r_offset);
get_rel_type_and_sym(elf, rela->r_info, &r_type, &r_sym);
tsym = elf->symtab_start + r_sym;
taddr = tsym->st_value + TO_NATIVE(rela->r_addend);
switch (elf->hdr->e_machine) {
case EM_RISCV:
if (!strcmp("__ex_table", fromsec) &&
r_type == R_RISCV_SUB32)
continue;
break;
case EM_LOONGARCH:
switch (r_type) {
case R_LARCH_SUB32:
if (!strcmp("__ex_table", fromsec))
continue;
break;
case R_LARCH_RELAX:
case R_LARCH_ALIGN:
/* These relocs do not refer to symbols */
continue;
}
break;
}
check_section_mismatch(mod, elf, tsym,
fsecndx, fromsec, r_offset, taddr);
}
}
static void section_rel(struct module *mod, struct elf_info *elf,
unsigned int fsecndx, const char *fromsec,
const Elf_Rel *start, const Elf_Rel *stop)
{
const Elf_Rel *rel;
for (rel = start; rel < stop; rel++) {
Elf_Sym *tsym;
Elf_Addr taddr, r_offset;
unsigned int r_type, r_sym;
void *loc;
r_offset = TO_NATIVE(rel->r_offset);
get_rel_type_and_sym(elf, rel->r_info, &r_type, &r_sym);
loc = sym_get_data_by_offset(elf, fsecndx, r_offset);
tsym = elf->symtab_start + r_sym;
switch (elf->hdr->e_machine) {
case EM_386:
taddr = addend_386_rel(loc, r_type);
break;
case EM_ARM:
taddr = addend_arm_rel(loc, tsym, r_type);
break;
case EM_MIPS:
taddr = addend_mips_rel(loc, r_type);
break;
default:
fatal("Please add code to calculate addend for this architecture\n");
}
check_section_mismatch(mod, elf, tsym,
fsecndx, fromsec, r_offset, taddr);
}
}
/**
* A module includes a number of sections that are discarded
* either when loaded or when used as built-in.
* For loaded modules all functions marked __init and all data
* marked __initdata will be discarded when the module has been initialized.
* Likewise for modules used built-in the sections marked __exit
* are discarded because __exit marked function are supposed to be called
* only when a module is unloaded which never happens for built-in modules.
* The check_sec_ref() function traverses all relocation records
* to find all references to a section that reference a section that will
* be discarded and warns about it.
**/
static void check_sec_ref(struct module *mod, struct elf_info *elf)
{
int i;
/* Walk through all sections */
for (i = 0; i < elf->num_sections; i++) {
Elf_Shdr *sechdr = &elf->sechdrs[i];
check_section(mod->name, elf, sechdr);
/* We want to process only relocation sections and not .init */
if (sechdr->sh_type == SHT_REL || sechdr->sh_type == SHT_RELA) {
/* section to which the relocation applies */
unsigned int secndx = sechdr->sh_info;
const char *secname = sec_name(elf, secndx);
const void *start, *stop;
/* If the section is known good, skip it */
if (match(secname, section_white_list))
continue;
start = sym_get_data_by_offset(elf, i, 0);
stop = start + sechdr->sh_size;
if (sechdr->sh_type == SHT_RELA)
section_rela(mod, elf, secndx, secname,
start, stop);
else
section_rel(mod, elf, secndx, secname,
start, stop);
}
}
}
static char *remove_dot(char *s)
{
size_t n = strcspn(s, ".");
if (n && s[n]) {
size_t m = strspn(s + n + 1, "0123456789");
if (m && (s[n + m + 1] == '.' || s[n + m + 1] == 0))
s[n] = 0;
}
return s;
}
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
/*
* The CRCs are recorded in .*.cmd files in the form of:
* #SYMVER <name> <crc>
*/
static void extract_crcs_for_object(const char *object, struct module *mod)
{
char cmd_file[PATH_MAX];
char *buf, *p;
const char *base;
int dirlen, ret;
base = strrchr(object, '/');
if (base) {
base++;
dirlen = base - object;
} else {
dirlen = 0;
base = object;
}
ret = snprintf(cmd_file, sizeof(cmd_file), "%.*s.%s.cmd",
dirlen, object, base);
if (ret >= sizeof(cmd_file)) {
error("%s: too long path was truncated\n", cmd_file);
return;
}
buf = read_text_file(cmd_file);
p = buf;
while ((p = strstr(p, "\n#SYMVER "))) {
char *name;
size_t namelen;
unsigned int crc;
struct symbol *sym;
name = p + strlen("\n#SYMVER ");
p = strchr(name, ' ');
if (!p)
break;
namelen = p - name;
p++;
if (!isdigit(*p))
continue; /* skip this line */
crc = strtoul(p, &p, 0);
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
if (*p != '\n')
continue; /* skip this line */
name[namelen] = '\0';
/*
* sym_find_with_module() may return NULL here.
* It typically occurs when CONFIG_TRIM_UNUSED_KSYMS=y.
* Since commit e1327a127703, genksyms calculates CRCs of all
* symbols, including trimmed ones. Ignore orphan CRCs.
*/
sym = sym_find_with_module(name, mod);
if (sym)
sym_set_crc(sym, crc);
}
free(buf);
}
/*
* The symbol versions (CRC) are recorded in the .*.cmd files.
* Parse them to retrieve CRCs for the current module.
*/
static void mod_set_crcs(struct module *mod)
{
char objlist[PATH_MAX];
char *buf, *p, *obj;
int ret;
if (mod->is_vmlinux) {
strcpy(objlist, ".vmlinux.objs");
} else {
/* objects for a module are listed in the *.mod file. */
ret = snprintf(objlist, sizeof(objlist), "%s.mod", mod->name);
if (ret >= sizeof(objlist)) {
error("%s: too long path was truncated\n", objlist);
return;
}
}
buf = read_text_file(objlist);
p = buf;
while ((obj = strsep(&p, "\n")) && obj[0])
extract_crcs_for_object(obj, mod);
free(buf);
}
static void read_symbols(const char *modname)
{
const char *symname;
char *version;
char *license;
char *namespace;
struct module *mod;
struct elf_info info = { };
Elf_Sym *sym;
if (!parse_elf(&info, modname))
return;
if (!strends(modname, ".o")) {
error("%s: filename must be suffixed with .o\n", modname);
return;
}
/* strip trailing .o */
mod = new_module(modname, strlen(modname) - strlen(".o"));
if (!mod->is_vmlinux) {
license = get_modinfo(&info, "license");
if (!license)
error("missing MODULE_LICENSE() in %s\n", modname);
while (license) {
if (!license_is_gpl_compatible(license)) {
mod->is_gpl_compatible = false;
break;
}
license = get_next_modinfo(&info, "license", license);
}
namespace = get_modinfo(&info, "import_ns");
while (namespace) {
add_namespace(&mod->imported_namespaces, namespace);
namespace = get_next_modinfo(&info, "import_ns",
namespace);
}
if (extra_warn && !get_modinfo(&info, "description"))
warn("missing MODULE_DESCRIPTION() in %s\n", modname);
}
for (sym = info.symtab_start; sym < info.symtab_stop; sym++) {
symname = remove_dot(info.strtab + sym->st_name);
handle_symbol(mod, &info, sym, symname);
handle_moddevtable(mod, &info, sym, symname);
}
check_sec_ref(mod, &info);
if (!mod->is_vmlinux) {
version = get_modinfo(&info, "version");
if (version || all_versions)
get_src_version(mod->name, mod->srcversion,
sizeof(mod->srcversion) - 1);
}
parse_elf_finish(&info);
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
if (modversions) {
/*
* Our trick to get versioning for module struct etc. - it's
* never passed as an argument to an exported function, so
* the automatic versioning doesn't pick it up, but it's really
* important anyhow.
*/
sym_add_unresolved("module_layout", mod, false);
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
mod_set_crcs(mod);
}
}
static void read_symbols_from_files(const char *filename)
{
FILE *in = stdin;
char fname[PATH_MAX];
in = fopen(filename, "r");
if (!in)
fatal("Can't open filenames file %s: %m", filename);
while (fgets(fname, PATH_MAX, in) != NULL) {
if (strends(fname, "\n"))
fname[strlen(fname)-1] = '\0';
read_symbols(fname);
}
fclose(in);
}
#define SZ 500
/* We first write the generated file into memory using the
* following helper, then compare to the file on disk and
* only update the later if anything changed */
void __attribute__((format(printf, 2, 3))) buf_printf(struct buffer *buf,
const char *fmt, ...)
{
char tmp[SZ];
int len;
va_list ap;
va_start(ap, fmt);
len = vsnprintf(tmp, SZ, fmt, ap);
buf_write(buf, tmp, len);
va_end(ap);
}
void buf_write(struct buffer *buf, const char *s, int len)
{
if (buf->size - buf->pos < len) {
buf->size += len + SZ;
buf->p = NOFAIL(realloc(buf->p, buf->size));
}
strncpy(buf->p + buf->pos, s, len);
buf->pos += len;
}
static void check_exports(struct module *mod)
{
struct symbol *s, *exp;
list_for_each_entry(s, &mod->unresolved_symbols, list) {
const char *basename;
exp = find_symbol(s->name);
if (!exp) {
kbuild: fix false-positive modpost warning when all symbols are trimmed Nathan reports that the mips defconfig emits the following warning: WARNING: modpost: Symbol info of vmlinux is missing. Unresolved symbol check will be entirely skipped. This false-positive happens when CONFIG_TRIM_UNUSED_KSYMS is enabled, but no CONFIG option is set to 'm'. Commit a0590473c5e6 ("nfs: fix PNFS_FLEXFILE_LAYOUT Kconfig default") turned the last 'm' into 'y' for the mips defconfig, and uncovered this issue. In this case, the module feature itself is enabled, but we have no module to build. As a result, CONFIG_TRIM_UNUSED_KSYMS drops all the instances of EXPORT_SYMBOL. Then, modpost wrongly assumes vmlinux is missing because vmlinux.symvers is empty. (As another false-positive case, you can create a module that does not use any symbol of vmlinux). The current behavior is to entirely suppress the unresolved symbol warnings when vmlinux is missing just because there are too many. I found the origin of this code in the historical git tree. [1] If this is a matter of noisiness, I think modpost can display the first 10 warnings, and the number of suppressed warnings at the end. You will get a bit noisier logs when you run 'make modules' without vmlinux, but such warnings are better to show because you never know the resulting modules are actually loadable or not. This commit changes the following: - If any of input *.symver files is missing, pass -w option to let the module build keep going with warnings instead of errors. - If there are too many (10+) unresolved symbol warnings, show only the first 10, and also the number of suppressed warnings. [1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=1cc0e0529569bf6a94f6d49770aa6d4b599d2c46 Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2021-03-25 18:54:11 +00:00
if (!s->weak && nr_unresolved++ < MAX_UNRESOLVED_REPORTS)
modpost_log(warn_unresolved ? LOG_WARN : LOG_ERROR,
"\"%s\" [%s.ko] undefined!\n",
s->name, mod->name);
continue;
}
if (exp->module == mod) {
error("\"%s\" [%s.ko] was exported without definition\n",
s->name, mod->name);
continue;
}
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
exp->used = true;
s->module = exp->module;
s->crc_valid = exp->crc_valid;
s->crc = exp->crc;
basename = strrchr(mod->name, '/');
if (basename)
basename++;
else
basename = mod->name;
if (!contains_namespace(&mod->imported_namespaces, exp->namespace)) {
modpost_log(allow_missing_ns_imports ? LOG_WARN : LOG_ERROR,
"module %s uses symbol %s from namespace %s, but does not import it.\n",
basename, exp->name, exp->namespace);
add_namespace(&mod->missing_namespaces, exp->namespace);
}
modpost: change the license of EXPORT_SYMBOL to bool type There were more EXPORT_SYMBOL types in the past. The following commits removed unused ones. - f1c3d73e973c ("module: remove EXPORT_SYMBOL_GPL_FUTURE") - 367948220fce ("module: remove EXPORT_UNUSED_SYMBOL*") There are 3 remaining in enum export, but export_unknown does not make any sense because we never expect such a situation like "we do not know how it was exported". If the symbol name starts with "__ksymtab_", but the section name does not start with "___ksymtab+" or "___ksymtab_gpl+", it is not an exported symbol. It occurs when a variable starting with "__ksymtab_" is directly defined: int __ksymtab_foo; Presumably, there is no practical issue for using such a weird variable name (but there is no good reason for doing so, either). Anyway, that is not an exported symbol. Setting export_unknown is not the right thing to do. Do not call sym_add_exported() in this case. With pointless export_unknown removed, the export type finally becomes boolean (either EXPORT_SYMBOL or EXPORT_SYMBOL_GPL). I renamed the field name to is_gpl_only. EXPORT_SYMBOL_GPL sets it true. Only GPL-compatible modules can use it. I removed the orphan comment, "How a symbol is exported", which is unrelated to sec_mismatch_count. It is about enum export. See commit bd5cbcedf446 ("kbuild: export-type enhancement to modpost.c") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org>
2022-05-08 19:06:19 +00:00
if (!mod->is_gpl_compatible && exp->is_gpl_only)
error("GPL-incompatible module %s.ko uses GPL-only symbol '%s'\n",
basename, exp->name);
}
}
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
static void handle_white_list_exports(const char *white_list)
{
char *buf, *p, *name;
buf = read_text_file(white_list);
p = buf;
while ((name = strsep(&p, "\n"))) {
struct symbol *sym = find_symbol(name);
if (sym)
sym->used = true;
}
free(buf);
}
static void check_modname_len(struct module *mod)
{
const char *mod_name;
mod_name = strrchr(mod->name, '/');
if (mod_name == NULL)
mod_name = mod->name;
else
mod_name++;
if (strlen(mod_name) >= MODULE_NAME_LEN)
error("module name is too long [%s.ko]\n", mod->name);
}
/**
* Header for the generated file
**/
static void add_header(struct buffer *b, struct module *mod)
{
buf_printf(b, "#include <linux/module.h>\n");
/*
* Include build-salt.h after module.h in order to
* inherit the definitions.
*/
buf_printf(b, "#define INCLUDE_VERMAGIC\n");
buf_printf(b, "#include <linux/build-salt.h>\n");
kbuild: add an elfnote for whether vmlinux is built with lto Currently, clang LTO built vmlinux won't work with pahole. LTO introduced cross-cu dwarf tag references and broke current pahole model which handles one cu as a time. The solution is to merge all cu's as one pahole cu as in [1]. We would like to do this merging only if cross-cu dwarf references happens. The LTO build mode is a pretty good indication for that. In earlier version of this patch ([2]), clang flag -grecord-gcc-switches is proposed to add to compilation flags so pahole could detect "-flto" and then merging cu's. This will increate the binary size of 1% without LTO though. Arnaldo suggested to use a note to indicate the vmlinux is built with LTO. Such a cheap way to get whether the vmlinux is built with LTO or not helps pahole but is also useful for tracing as LTO may inline/delete/demote global functions, promote static functions, etc. So this patch added an elfnote with a new type LINUX_ELFNOTE_LTO_INFO. The owner of the note is "Linux". With gcc 8.4.1 and clang trunk, without LTO, I got $ readelf -n vmlinux Displaying notes found in: .notes Owner Data size Description ... Linux 0x00000004 func description data: 00 00 00 00 ... With "readelf -x ".notes" vmlinux", I can verify the above "func" with type code 0x101. With clang thin-LTO, I got the same as above except the following: description data: 01 00 00 00 which indicates the vmlinux is built with LTO. [1] https://lore.kernel.org/bpf/20210325065316.3121287-1-yhs@fb.com/ [2] https://lore.kernel.org/bpf/20210331001623.2778934-1-yhs@fb.com/ Suggested-by: Arnaldo Carvalho de Melo <arnaldo.melo@gmail.com> Signed-off-by: Yonghong Song <yhs@fb.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM/Clang v12.0.0-rc4 (x86-64) Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2021-04-01 23:27:23 +00:00
buf_printf(b, "#include <linux/elfnote-lto.h>\n");
2022-05-13 11:39:22 +00:00
buf_printf(b, "#include <linux/export-internal.h>\n");
buf_printf(b, "#include <linux/vermagic.h>\n");
buf_printf(b, "#include <linux/compiler.h>\n");
buf_printf(b, "\n");
x86/unwind/orc: Add ELF section with ORC version identifier Commits ffb1b4a41016 ("x86/unwind/orc: Add 'signal' field to ORC metadata") and fb799447ae29 ("x86,objtool: Split UNWIND_HINT_EMPTY in two") changed the ORC format. Although ORC is internal to the kernel, it's the only way for external tools to get reliable kernel stack traces on x86-64. In particular, the drgn debugger [1] uses ORC for stack unwinding, and these format changes broke it [2]. As the drgn maintainer, I don't care how often or how much the kernel changes the ORC format as long as I have a way to detect the change. It suffices to store a version identifier in the vmlinux and kernel module ELF files (to use when parsing ORC sections from ELF), and in kernel memory (to use when parsing ORC from a core dump+symbol table). Rather than hard-coding a version number that needs to be manually bumped, Peterz suggested hashing the definitions from orc_types.h. If there is a format change that isn't caught by this, the hashing script can be updated. This patch adds an .orc_header allocated ELF section containing the 20-byte hash to vmlinux and kernel modules, along with the corresponding __start_orc_header and __stop_orc_header symbols in vmlinux. 1: https://github.com/osandov/drgn 2: https://github.com/osandov/drgn/issues/303 Fixes: ffb1b4a41016 ("x86/unwind/orc: Add 'signal' field to ORC metadata") Fixes: fb799447ae29 ("x86,objtool: Split UNWIND_HINT_EMPTY in two") Signed-off-by: Omar Sandoval <osandov@fb.com> Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org> Acked-by: Josh Poimboeuf <jpoimboe@kernel.org> Link: https://lkml.kernel.org/r/aef9c8dc43915b886a8c48509a12ec1b006ca1ca.1686690801.git.osandov@osandov.com
2023-06-13 21:14:56 +00:00
buf_printf(b, "#ifdef CONFIG_UNWINDER_ORC\n");
buf_printf(b, "#include <asm/orc_header.h>\n");
buf_printf(b, "ORC_HEADER;\n");
buf_printf(b, "#endif\n");
buf_printf(b, "\n");
buf_printf(b, "BUILD_SALT;\n");
kbuild: add an elfnote for whether vmlinux is built with lto Currently, clang LTO built vmlinux won't work with pahole. LTO introduced cross-cu dwarf tag references and broke current pahole model which handles one cu as a time. The solution is to merge all cu's as one pahole cu as in [1]. We would like to do this merging only if cross-cu dwarf references happens. The LTO build mode is a pretty good indication for that. In earlier version of this patch ([2]), clang flag -grecord-gcc-switches is proposed to add to compilation flags so pahole could detect "-flto" and then merging cu's. This will increate the binary size of 1% without LTO though. Arnaldo suggested to use a note to indicate the vmlinux is built with LTO. Such a cheap way to get whether the vmlinux is built with LTO or not helps pahole but is also useful for tracing as LTO may inline/delete/demote global functions, promote static functions, etc. So this patch added an elfnote with a new type LINUX_ELFNOTE_LTO_INFO. The owner of the note is "Linux". With gcc 8.4.1 and clang trunk, without LTO, I got $ readelf -n vmlinux Displaying notes found in: .notes Owner Data size Description ... Linux 0x00000004 func description data: 00 00 00 00 ... With "readelf -x ".notes" vmlinux", I can verify the above "func" with type code 0x101. With clang thin-LTO, I got the same as above except the following: description data: 01 00 00 00 which indicates the vmlinux is built with LTO. [1] https://lore.kernel.org/bpf/20210325065316.3121287-1-yhs@fb.com/ [2] https://lore.kernel.org/bpf/20210331001623.2778934-1-yhs@fb.com/ Suggested-by: Arnaldo Carvalho de Melo <arnaldo.melo@gmail.com> Signed-off-by: Yonghong Song <yhs@fb.com> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM/Clang v12.0.0-rc4 (x86-64) Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2021-04-01 23:27:23 +00:00
buf_printf(b, "BUILD_LTO_INFO;\n");
buf_printf(b, "\n");
buf_printf(b, "MODULE_INFO(vermagic, VERMAGIC_STRING);\n");
buf_printf(b, "MODULE_INFO(name, KBUILD_MODNAME);\n");
buf_printf(b, "\n");
buf_printf(b, "__visible struct module __this_module\n");
buf_printf(b, "__section(\".gnu.linkonce.this_module\") = {\n");
buf_printf(b, "\t.name = KBUILD_MODNAME,\n");
if (mod->has_init)
buf_printf(b, "\t.init = init_module,\n");
if (mod->has_cleanup)
buf_printf(b, "#ifdef CONFIG_MODULE_UNLOAD\n"
"\t.exit = cleanup_module,\n"
"#endif\n");
buf_printf(b, "\t.arch = MODULE_ARCH_INIT,\n");
buf_printf(b, "};\n");
if (!external_module)
buf_printf(b, "\nMODULE_INFO(intree, \"Y\");\n");
buf_printf(b,
"\n"
"#ifdef CONFIG_MITIGATION_RETPOLINE\n"
"MODULE_INFO(retpoline, \"Y\");\n"
"#endif\n");
if (strstarts(mod->name, "drivers/staging"))
buf_printf(b, "\nMODULE_INFO(staging, \"Y\");\n");
if (strstarts(mod->name, "tools/testing"))
buf_printf(b, "\nMODULE_INFO(test, \"Y\");\n");
}
2022-05-13 11:39:22 +00:00
static void add_exported_symbols(struct buffer *buf, struct module *mod)
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
{
struct symbol *sym;
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
/* generate struct for exported symbols */
buf_printf(buf, "\n");
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
list_for_each_entry(sym, &mod->exported_symbols, list) {
if (trim_unused_exports && !sym->used)
continue;
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
buf_printf(buf, "KSYMTAB_%s(%s, \"%s\", \"%s\");\n",
sym->is_func ? "FUNC" : "DATA", sym->name,
sym->is_gpl_only ? "_gpl" : "", sym->namespace);
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
}
kbuild: generate KSYMTAB entries by modpost Commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") made modpost output CRCs in the same way whether the EXPORT_SYMBOL() is placed in *.c or *.S. For further cleanups, this commit applies a similar approach to the entire data structure of EXPORT_SYMBOL(). The EXPORT_SYMBOL() compilation is split into two stages. When a source file is compiled, EXPORT_SYMBOL() will be converted into a dummy symbol in the .export_symbol section. For example, EXPORT_SYMBOL(foo); EXPORT_SYMBOL_NS_GPL(bar, BAR_NAMESPACE); will be encoded into the following assembly code: .section ".export_symbol","a" __export_symbol_foo: .asciz "" /* license */ .asciz "" /* name space */ .balign 8 .quad foo /* symbol reference */ .previous .section ".export_symbol","a" __export_symbol_bar: .asciz "GPL" /* license */ .asciz "BAR_NAMESPACE" /* name space */ .balign 8 .quad bar /* symbol reference */ .previous They are mere markers to tell modpost the name, license, and namespace of the symbols. They will be dropped from the final vmlinux and modules because the *(.export_symbol) will go into /DISCARD/ in the linker script. Then, modpost extracts all the information about EXPORT_SYMBOL() from the .export_symbol section, and generates the final C code: KSYMTAB_FUNC(foo, "", ""); KSYMTAB_FUNC(bar, "_gpl", "BAR_NAMESPACE"); KSYMTAB_FUNC() (or KSYMTAB_DATA() if it is data) is expanded to struct kernel_symbol that will be linked to the vmlinux or a module. With this change, EXPORT_SYMBOL() works in the same way for *.c and *.S files, providing the following benefits. [1] Deprecate EXPORT_DATA_SYMBOL() In the old days, EXPORT_SYMBOL() was only available in C files. To export a symbol in *.S, EXPORT_SYMBOL() was placed in a separate *.c file. arch/arm/kernel/armksyms.c is one example written in the classic manner. Commit 22823ab419d8 ("EXPORT_SYMBOL() for asm") removed this limitation. Since then, EXPORT_SYMBOL() can be placed close to the symbol definition in *.S files. It was a nice improvement. However, as that commit mentioned, you need to use EXPORT_DATA_SYMBOL() for data objects on some architectures. In the new approach, modpost checks symbol's type (STT_FUNC or not), and outputs KSYMTAB_FUNC() or KSYMTAB_DATA() accordingly. There are only two users of EXPORT_DATA_SYMBOL: EXPORT_DATA_SYMBOL_GPL(empty_zero_page) (arch/ia64/kernel/head.S) EXPORT_DATA_SYMBOL(ia64_ivt) (arch/ia64/kernel/ivt.S) They are transformed as follows and output into .vmlinux.export.c KSYMTAB_DATA(empty_zero_page, "_gpl", ""); KSYMTAB_DATA(ia64_ivt, "", ""); The other EXPORT_SYMBOL users in ia64 assembly are output as KSYMTAB_FUNC(). EXPORT_DATA_SYMBOL() is now deprecated. [2] merge <linux/export.h> and <asm-generic/export.h> There are two similar header implementations: include/linux/export.h for .c files include/asm-generic/export.h for .S files Ideally, the functionality should be consistent between them, but they tend to diverge. Commit 8651ec01daed ("module: add support for symbol namespaces.") did not support the namespace for *.S files. This commit shifts the essential implementation part to C, which supports EXPORT_SYMBOL_NS() for *.S files. <asm/export.h> and <asm-generic/export.h> will remain as a wrapper of <linux/export.h> for a while. They will be removed after #include <asm/export.h> directives are all replaced with #include <linux/export.h>. [3] Implement CONFIG_TRIM_UNUSED_KSYMS in one-pass algorithm (by a later commit) When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. We can do this better now; modpost can selectively emit KSYMTAB entries that are really used by modules. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
2023-06-11 15:50:52 +00:00
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
if (!modversions)
return;
2022-05-13 11:39:22 +00:00
/* record CRCs for exported symbols */
buf_printf(buf, "\n");
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
list_for_each_entry(sym, &mod->exported_symbols, list) {
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
if (trim_unused_exports && !sym->used)
continue;
modpost: fix module versioning when a symbol lacks valid CRC Since commit 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS"), module versioning is broken on some architectures. Loading a module fails with "disagrees about version of symbol module_layout". On such architectures (e.g. ARCH=sparc build with sparc64_defconfig), modpost shows a warning, like follows: WARNING: modpost: EXPORT symbol "_mcount" [vmlinux] version generation failed, symbol will not be versioned. Is "_mcount" prototyped in <asm/asm-prototypes.h>? Previously, it was a harmless warning (CRC check was just skipped), but now wrong CRCs are used for comparison because invalid CRCs are just skipped. $ sparc64-linux-gnu-nm -n vmlinux [snip] 0000000000c2cea0 r __ksymtab__kstrtol 0000000000c2ceb8 r __ksymtab__kstrtoul 0000000000c2ced0 r __ksymtab__local_bh_enable 0000000000c2cee8 r __ksymtab__mcount 0000000000c2cf00 r __ksymtab__printk 0000000000c2cf18 r __ksymtab__raw_read_lock 0000000000c2cf30 r __ksymtab__raw_read_lock_bh [snip] 0000000000c53b34 D __crc__kstrtol 0000000000c53b38 D __crc__kstrtoul 0000000000c53b3c D __crc__local_bh_enable 0000000000c53b40 D __crc__printk 0000000000c53b44 D __crc__raw_read_lock 0000000000c53b48 D __crc__raw_read_lock_bh Please notice __crc__mcount is missing here. When the module subsystem looks up a CRC that comes after, it results in reading out a wrong address. For example, when __crc__printk is needed, the module subsystem reads 0xc53b44 instead of 0xc53b40. All CRC entries must be output for correct index accessing. Invalid CRCs will be unused, but are needed to keep the one-to-one mapping between __ksymtab_* and __crc_*. The best is to fix all modpost warnings, but several warnings are still remaining on less popular architectures. Fixes: 7b4537199a4a ("kbuild: link symbol CRCs at final link, removing CONFIG_MODULE_REL_CRCS") Reported-by: matoro <matoro_mailinglist_kernel@matoro.tk> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Tested-by: matoro <matoro_mailinglist_kernel@matoro.tk>
2022-08-09 14:11:17 +00:00
if (!sym->crc_valid)
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
warn("EXPORT symbol \"%s\" [%s%s] version generation failed, symbol will not be versioned.\n"
"Is \"%s\" prototyped in <asm/asm-prototypes.h>?\n",
sym->name, mod->name, mod->is_vmlinux ? "" : ".ko",
sym->name);
2022-05-13 11:39:22 +00:00
buf_printf(buf, "SYMBOL_CRC(%s, 0x%08x, \"%s\");\n",
sym->name, sym->crc, sym->is_gpl_only ? "_gpl" : "");
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
}
}
/**
* Record CRCs for unresolved symbols
**/
static void add_versions(struct buffer *b, struct module *mod)
{
struct symbol *s;
if (!modversions)
return;
buf_printf(b, "\n");
buf_printf(b, "static const struct modversion_info ____versions[]\n");
buf_printf(b, "__used __section(\"__versions\") = {\n");
list_for_each_entry(s, &mod->unresolved_symbols, list) {
if (!s->module)
continue;
if (!s->crc_valid) {
warn("\"%s\" [%s.ko] has no CRC!\n",
s->name, mod->name);
continue;
}
if (strlen(s->name) >= MODULE_NAME_LEN) {
error("too long symbol \"%s\" [%s.ko]\n",
s->name, mod->name);
break;
}
buf_printf(b, "\t{ %#8x, \"%s\" },\n",
s->crc, s->name);
}
buf_printf(b, "};\n");
}
static void add_depends(struct buffer *b, struct module *mod)
{
struct symbol *s;
int first = 1;
/* Clear ->seen flag of modules that own symbols needed by this. */
list_for_each_entry(s, &mod->unresolved_symbols, list) {
if (s->module)
s->module->seen = s->module->is_vmlinux;
}
buf_printf(b, "\n");
buf_printf(b, "MODULE_INFO(depends, \"");
list_for_each_entry(s, &mod->unresolved_symbols, list) {
const char *p;
if (!s->module)
continue;
if (s->module->seen)
continue;
s->module->seen = true;
p = strrchr(s->module->name, '/');
if (p)
p++;
else
p = s->module->name;
buf_printf(b, "%s%s", first ? "" : ",", p);
first = 0;
}
buf_printf(b, "\");\n");
}
static void add_srcversion(struct buffer *b, struct module *mod)
{
if (mod->srcversion[0]) {
buf_printf(b, "\n");
buf_printf(b, "MODULE_INFO(srcversion, \"%s\");\n",
mod->srcversion);
}
}
static void write_buf(struct buffer *b, const char *fname)
{
FILE *file;
if (error_occurred)
return;
file = fopen(fname, "w");
if (!file) {
perror(fname);
exit(1);
}
if (fwrite(b->p, 1, b->pos, file) != b->pos) {
perror(fname);
exit(1);
}
if (fclose(file) != 0) {
perror(fname);
exit(1);
}
}
static void write_if_changed(struct buffer *b, const char *fname)
{
char *tmp;
FILE *file;
struct stat st;
file = fopen(fname, "r");
if (!file)
goto write;
if (fstat(fileno(file), &st) < 0)
goto close_write;
if (st.st_size != b->pos)
goto close_write;
tmp = NOFAIL(malloc(b->pos));
if (fread(tmp, 1, b->pos, file) != b->pos)
goto free_write;
if (memcmp(tmp, b->p, b->pos) != 0)
goto free_write;
free(tmp);
fclose(file);
return;
free_write:
free(tmp);
close_write:
fclose(file);
write:
write_buf(b, fname);
}
2022-05-13 11:39:22 +00:00
static void write_vmlinux_export_c_file(struct module *mod)
{
struct buffer buf = { };
buf_printf(&buf,
"#include <linux/export-internal.h>\n");
add_exported_symbols(&buf, mod);
write_if_changed(&buf, ".vmlinux.export.c");
free(buf.p);
}
/* do sanity checks, and generate *.mod.c file */
static void write_mod_c_file(struct module *mod)
{
struct buffer buf = { };
char fname[PATH_MAX];
int ret;
add_header(&buf, mod);
2022-05-13 11:39:22 +00:00
add_exported_symbols(&buf, mod);
add_versions(&buf, mod);
add_depends(&buf, mod);
add_moddevtable(&buf, mod);
add_srcversion(&buf, mod);
ret = snprintf(fname, sizeof(fname), "%s.mod.c", mod->name);
if (ret >= sizeof(fname)) {
error("%s: too long path was truncated\n", fname);
goto free;
}
write_if_changed(&buf, fname);
free:
free(buf.p);
}
/* parse Module.symvers file. line format:
modpost: move the namespace field in Module.symvers last In order to preserve backwards compatability with kmod tools, we have to move the namespace field in Module.symvers last, as the depmod -e -E option looks at the first three fields in Module.symvers to check symbol versions (and it's expected they stay in the original order of crc, symbol, module). In addition, update an ancient comment above read_dump() in modpost that suggested that the export type field in Module.symvers was optional. I suspect that there were historical reasons behind that comment that are no longer accurate. We have been unconditionally printing the export type since 2.6.18 (commit bd5cbcedf44), which is over a decade ago now. Fix up read_dump() to treat each field as non-optional. I suspect the original read_dump() code treated the export field as optional in order to support pre <= 2.6.18 Module.symvers (which did not have the export type field). Note that although symbol namespaces are optional, the field will not be omitted from Module.symvers if a symbol does not have a namespace. In this case, the field will simply be empty and the next delimiter or end of line will follow. Cc: stable@vger.kernel.org Fixes: cb9b55d21fe0 ("modpost: add support for symbol namespaces") Tested-by: Matthias Maennich <maennich@google.com> Reviewed-by: Matthias Maennich <maennich@google.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2020-03-11 17:01:20 +00:00
* 0x12345678<tab>symbol<tab>module<tab>export<tab>namespace
**/
static void read_dump(const char *fname)
{
char *buf, *pos, *line;
buf = read_text_file(fname);
if (!buf)
/* No symbol versions, silently ignore */
return;
pos = buf;
while ((line = get_line(&pos))) {
modpost: move the namespace field in Module.symvers last In order to preserve backwards compatability with kmod tools, we have to move the namespace field in Module.symvers last, as the depmod -e -E option looks at the first three fields in Module.symvers to check symbol versions (and it's expected they stay in the original order of crc, symbol, module). In addition, update an ancient comment above read_dump() in modpost that suggested that the export type field in Module.symvers was optional. I suspect that there were historical reasons behind that comment that are no longer accurate. We have been unconditionally printing the export type since 2.6.18 (commit bd5cbcedf44), which is over a decade ago now. Fix up read_dump() to treat each field as non-optional. I suspect the original read_dump() code treated the export field as optional in order to support pre <= 2.6.18 Module.symvers (which did not have the export type field). Note that although symbol namespaces are optional, the field will not be omitted from Module.symvers if a symbol does not have a namespace. In this case, the field will simply be empty and the next delimiter or end of line will follow. Cc: stable@vger.kernel.org Fixes: cb9b55d21fe0 ("modpost: add support for symbol namespaces") Tested-by: Matthias Maennich <maennich@google.com> Reviewed-by: Matthias Maennich <maennich@google.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2020-03-11 17:01:20 +00:00
char *symname, *namespace, *modname, *d, *export;
unsigned int crc;
struct module *mod;
struct symbol *s;
modpost: change the license of EXPORT_SYMBOL to bool type There were more EXPORT_SYMBOL types in the past. The following commits removed unused ones. - f1c3d73e973c ("module: remove EXPORT_SYMBOL_GPL_FUTURE") - 367948220fce ("module: remove EXPORT_UNUSED_SYMBOL*") There are 3 remaining in enum export, but export_unknown does not make any sense because we never expect such a situation like "we do not know how it was exported". If the symbol name starts with "__ksymtab_", but the section name does not start with "___ksymtab+" or "___ksymtab_gpl+", it is not an exported symbol. It occurs when a variable starting with "__ksymtab_" is directly defined: int __ksymtab_foo; Presumably, there is no practical issue for using such a weird variable name (but there is no good reason for doing so, either). Anyway, that is not an exported symbol. Setting export_unknown is not the right thing to do. Do not call sym_add_exported() in this case. With pointless export_unknown removed, the export type finally becomes boolean (either EXPORT_SYMBOL or EXPORT_SYMBOL_GPL). I renamed the field name to is_gpl_only. EXPORT_SYMBOL_GPL sets it true. Only GPL-compatible modules can use it. I removed the orphan comment, "How a symbol is exported", which is unrelated to sec_mismatch_count. It is about enum export. See commit bd5cbcedf446 ("kbuild: export-type enhancement to modpost.c") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org>
2022-05-08 19:06:19 +00:00
bool gpl_only;
if (!(symname = strchr(line, '\t')))
goto fail;
*symname++ = '\0';
modpost: move the namespace field in Module.symvers last In order to preserve backwards compatability with kmod tools, we have to move the namespace field in Module.symvers last, as the depmod -e -E option looks at the first three fields in Module.symvers to check symbol versions (and it's expected they stay in the original order of crc, symbol, module). In addition, update an ancient comment above read_dump() in modpost that suggested that the export type field in Module.symvers was optional. I suspect that there were historical reasons behind that comment that are no longer accurate. We have been unconditionally printing the export type since 2.6.18 (commit bd5cbcedf44), which is over a decade ago now. Fix up read_dump() to treat each field as non-optional. I suspect the original read_dump() code treated the export field as optional in order to support pre <= 2.6.18 Module.symvers (which did not have the export type field). Note that although symbol namespaces are optional, the field will not be omitted from Module.symvers if a symbol does not have a namespace. In this case, the field will simply be empty and the next delimiter or end of line will follow. Cc: stable@vger.kernel.org Fixes: cb9b55d21fe0 ("modpost: add support for symbol namespaces") Tested-by: Matthias Maennich <maennich@google.com> Reviewed-by: Matthias Maennich <maennich@google.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2020-03-11 17:01:20 +00:00
if (!(modname = strchr(symname, '\t')))
goto fail;
*modname++ = '\0';
modpost: move the namespace field in Module.symvers last In order to preserve backwards compatability with kmod tools, we have to move the namespace field in Module.symvers last, as the depmod -e -E option looks at the first three fields in Module.symvers to check symbol versions (and it's expected they stay in the original order of crc, symbol, module). In addition, update an ancient comment above read_dump() in modpost that suggested that the export type field in Module.symvers was optional. I suspect that there were historical reasons behind that comment that are no longer accurate. We have been unconditionally printing the export type since 2.6.18 (commit bd5cbcedf44), which is over a decade ago now. Fix up read_dump() to treat each field as non-optional. I suspect the original read_dump() code treated the export field as optional in order to support pre <= 2.6.18 Module.symvers (which did not have the export type field). Note that although symbol namespaces are optional, the field will not be omitted from Module.symvers if a symbol does not have a namespace. In this case, the field will simply be empty and the next delimiter or end of line will follow. Cc: stable@vger.kernel.org Fixes: cb9b55d21fe0 ("modpost: add support for symbol namespaces") Tested-by: Matthias Maennich <maennich@google.com> Reviewed-by: Matthias Maennich <maennich@google.com> Reviewed-by: Lucas De Marchi <lucas.demarchi@intel.com> Signed-off-by: Jessica Yu <jeyu@kernel.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2020-03-11 17:01:20 +00:00
if (!(export = strchr(modname, '\t')))
goto fail;
*export++ = '\0';
if (!(namespace = strchr(export, '\t')))
goto fail;
*namespace++ = '\0';
crc = strtoul(line, &d, 16);
if (*symname == '\0' || *modname == '\0' || *d != '\0')
goto fail;
modpost: change the license of EXPORT_SYMBOL to bool type There were more EXPORT_SYMBOL types in the past. The following commits removed unused ones. - f1c3d73e973c ("module: remove EXPORT_SYMBOL_GPL_FUTURE") - 367948220fce ("module: remove EXPORT_UNUSED_SYMBOL*") There are 3 remaining in enum export, but export_unknown does not make any sense because we never expect such a situation like "we do not know how it was exported". If the symbol name starts with "__ksymtab_", but the section name does not start with "___ksymtab+" or "___ksymtab_gpl+", it is not an exported symbol. It occurs when a variable starting with "__ksymtab_" is directly defined: int __ksymtab_foo; Presumably, there is no practical issue for using such a weird variable name (but there is no good reason for doing so, either). Anyway, that is not an exported symbol. Setting export_unknown is not the right thing to do. Do not call sym_add_exported() in this case. With pointless export_unknown removed, the export type finally becomes boolean (either EXPORT_SYMBOL or EXPORT_SYMBOL_GPL). I renamed the field name to is_gpl_only. EXPORT_SYMBOL_GPL sets it true. Only GPL-compatible modules can use it. I removed the orphan comment, "How a symbol is exported", which is unrelated to sec_mismatch_count. It is about enum export. See commit bd5cbcedf446 ("kbuild: export-type enhancement to modpost.c") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org>
2022-05-08 19:06:19 +00:00
if (!strcmp(export, "EXPORT_SYMBOL_GPL")) {
gpl_only = true;
} else if (!strcmp(export, "EXPORT_SYMBOL")) {
gpl_only = false;
} else {
error("%s: unknown license %s. skip", symname, export);
continue;
}
mod = find_module(modname);
if (!mod) {
mod = new_module(modname, strlen(modname));
mod->from_dump = true;
}
s = sym_add_exported(symname, mod, gpl_only, namespace);
modpost: extract symbol versions from *.cmd files Currently, CONFIG_MODVERSIONS needs extra link to embed the symbol versions into ELF objects. Then, modpost extracts the version CRCs from them. The following figures show how it currently works, and how I am trying to change it. Current implementation ====================== |----------| embed CRC -------------------------->| final | $(CC) $(LD) / |---------| | link for | -----> *.o -------> *.o -->| modpost | | vmlinux | / / | |-- *.mod.c -->| or | / genksyms / |---------| | module | *.c ------> *.symversions |----------| Genksyms outputs the calculated CRCs in the form of linker script (*.symversions), which is used by $(LD) to update the object. If CONFIG_LTO_CLANG=y, the build process is much more complex. Embedding the CRCs is postponed until the LLVM bitcode is converted into ELF, creating another intermediate *.prelink.o. However, this complexity is unneeded. There is no reason why we must embed version CRCs in objects so early. There is final link stage for vmlinux (scripts/link-vmlinux.sh) and modules (scripts/Makefile.modfinal). We can link CRCs at the very last moment. New implementation ================== |----------| --------------------------------------->| final | $(CC) / |---------| | link for | -----> *.o ---->| | | vmlinux | / | modpost |--- .vmlinux.export.c -->| or | / genksyms | |--- *.mod.c ------------>| module | *.c ------> *.cmd -->|---------| |----------| Pass the symbol versions to modpost as separate text data, which are available in *.cmd files. This commit changes modpost to extract CRCs from *.cmd files instead of from ELF objects. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org> Reviewed-by: Sami Tolvanen <samitolvanen@google.com> Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # LLVM-14 (x86-64)
2022-05-13 11:39:21 +00:00
sym_set_crc(s, crc);
}
free(buf);
return;
fail:
free(buf);
fatal("parse error in symbol dump file\n");
}
static void write_dump(const char *fname)
{
struct buffer buf = { };
struct module *mod;
struct symbol *sym;
list_for_each_entry(mod, &modules, list) {
if (mod->from_dump)
continue;
list_for_each_entry(sym, &mod->exported_symbols, list) {
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
if (trim_unused_exports && !sym->used)
continue;
modpost: change the license of EXPORT_SYMBOL to bool type There were more EXPORT_SYMBOL types in the past. The following commits removed unused ones. - f1c3d73e973c ("module: remove EXPORT_SYMBOL_GPL_FUTURE") - 367948220fce ("module: remove EXPORT_UNUSED_SYMBOL*") There are 3 remaining in enum export, but export_unknown does not make any sense because we never expect such a situation like "we do not know how it was exported". If the symbol name starts with "__ksymtab_", but the section name does not start with "___ksymtab+" or "___ksymtab_gpl+", it is not an exported symbol. It occurs when a variable starting with "__ksymtab_" is directly defined: int __ksymtab_foo; Presumably, there is no practical issue for using such a weird variable name (but there is no good reason for doing so, either). Anyway, that is not an exported symbol. Setting export_unknown is not the right thing to do. Do not call sym_add_exported() in this case. With pointless export_unknown removed, the export type finally becomes boolean (either EXPORT_SYMBOL or EXPORT_SYMBOL_GPL). I renamed the field name to is_gpl_only. EXPORT_SYMBOL_GPL sets it true. Only GPL-compatible modules can use it. I removed the orphan comment, "How a symbol is exported", which is unrelated to sec_mismatch_count. It is about enum export. See commit bd5cbcedf446 ("kbuild: export-type enhancement to modpost.c") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org>
2022-05-08 19:06:19 +00:00
buf_printf(&buf, "0x%08x\t%s\t%s\tEXPORT_SYMBOL%s\t%s\n",
sym->crc, sym->name, mod->name,
modpost: change the license of EXPORT_SYMBOL to bool type There were more EXPORT_SYMBOL types in the past. The following commits removed unused ones. - f1c3d73e973c ("module: remove EXPORT_SYMBOL_GPL_FUTURE") - 367948220fce ("module: remove EXPORT_UNUSED_SYMBOL*") There are 3 remaining in enum export, but export_unknown does not make any sense because we never expect such a situation like "we do not know how it was exported". If the symbol name starts with "__ksymtab_", but the section name does not start with "___ksymtab+" or "___ksymtab_gpl+", it is not an exported symbol. It occurs when a variable starting with "__ksymtab_" is directly defined: int __ksymtab_foo; Presumably, there is no practical issue for using such a weird variable name (but there is no good reason for doing so, either). Anyway, that is not an exported symbol. Setting export_unknown is not the right thing to do. Do not call sym_add_exported() in this case. With pointless export_unknown removed, the export type finally becomes boolean (either EXPORT_SYMBOL or EXPORT_SYMBOL_GPL). I renamed the field name to is_gpl_only. EXPORT_SYMBOL_GPL sets it true. Only GPL-compatible modules can use it. I removed the orphan comment, "How a symbol is exported", which is unrelated to sec_mismatch_count. It is about enum export. See commit bd5cbcedf446 ("kbuild: export-type enhancement to modpost.c") Signed-off-by: Masahiro Yamada <masahiroy@kernel.org> Reviewed-by: Nicolas Schier <nicolas@fjasle.eu> Tested-by: Nathan Chancellor <nathan@kernel.org>
2022-05-08 19:06:19 +00:00
sym->is_gpl_only ? "_GPL" : "",
sym->namespace);
}
}
write_buf(&buf, fname);
free(buf.p);
}
static void write_namespace_deps_files(const char *fname)
{
struct module *mod;
struct namespace_list *ns;
struct buffer ns_deps_buf = {};
list_for_each_entry(mod, &modules, list) {
if (mod->from_dump || list_empty(&mod->missing_namespaces))
continue;
buf_printf(&ns_deps_buf, "%s.ko:", mod->name);
list_for_each_entry(ns, &mod->missing_namespaces, list)
buf_printf(&ns_deps_buf, " %s", ns->namespace);
buf_printf(&ns_deps_buf, "\n");
}
write_if_changed(&ns_deps_buf, fname);
free(ns_deps_buf.p);
}
struct dump_list {
struct list_head list;
const char *file;
};
int main(int argc, char **argv)
{
struct module *mod;
char *missing_namespace_deps = NULL;
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
char *unused_exports_white_list = NULL;
char *dump_write = NULL, *files_source = NULL;
int opt;
LIST_HEAD(dump_lists);
struct dump_list *dl, *dl2;
while ((opt = getopt(argc, argv, "ei:MmnT:to:au:WwENd:")) != -1) {
switch (opt) {
case 'e':
external_module = true;
break;
case 'i':
dl = NOFAIL(malloc(sizeof(*dl)));
dl->file = optarg;
list_add_tail(&dl->list, &dump_lists);
break;
case 'M':
module_enabled = true;
break;
case 'm':
modversions = true;
break;
modpost: Optionally ignore secondary errors seen if a single module build fails Commit ea4054a23 (modpost: handle huge numbers of modules) added support for building a large number of modules. Unfortunately, the commit changed the semantics of the makefile: Instead of passing only existing object files to modpost, make now passes all expected object files. If make was started with option -i, this results in a modpost error if a single file failed to build. Example with the current btrfs build falure on m68k: fs/btrfs/btrfs.o: No such file or directory make[1]: [__modpost] Error 1 (ignored) This error is followed by lots of errors such as: m68k-linux-gcc: error: arch/m68k/emu/nfcon.mod.c: No such file or directory m68k-linux-gcc: fatal error: no input files compilation terminated. make[1]: [arch/m68k/emu/nfcon.mod.o] Error 1 (ignored) This doesn't matter much for normal builds, but it is annoying for builds started with "make -i" due to the large number of secondary errors. Those errors unnececessarily clog any error log and make it difficult to find the real errors in the build. Fix the problem by adding a new parameter '-n' to modpost. If this parameter is specified, modpost reports but ignores missing object files. With this patch, error output from above problem is (with make -i): m68k-linux-ld: cannot find fs/btrfs/ioctl.o: No such file or directory make[2]: [fs/btrfs/btrfs.o] Error 1 (ignored) ... fs/btrfs/btrfs.o: No such file or directory (ignored) Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Michael Marek <mmarek@suse.cz> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
2013-09-23 05:53:54 +00:00
case 'n':
ignore_missing_files = true;
modpost: Optionally ignore secondary errors seen if a single module build fails Commit ea4054a23 (modpost: handle huge numbers of modules) added support for building a large number of modules. Unfortunately, the commit changed the semantics of the makefile: Instead of passing only existing object files to modpost, make now passes all expected object files. If make was started with option -i, this results in a modpost error if a single file failed to build. Example with the current btrfs build falure on m68k: fs/btrfs/btrfs.o: No such file or directory make[1]: [__modpost] Error 1 (ignored) This error is followed by lots of errors such as: m68k-linux-gcc: error: arch/m68k/emu/nfcon.mod.c: No such file or directory m68k-linux-gcc: fatal error: no input files compilation terminated. make[1]: [arch/m68k/emu/nfcon.mod.o] Error 1 (ignored) This doesn't matter much for normal builds, but it is annoying for builds started with "make -i" due to the large number of secondary errors. Those errors unnececessarily clog any error log and make it difficult to find the real errors in the build. Fix the problem by adding a new parameter '-n' to modpost. If this parameter is specified, modpost reports but ignores missing object files. With this patch, error output from above problem is (with make -i): m68k-linux-ld: cannot find fs/btrfs/ioctl.o: No such file or directory make[2]: [fs/btrfs/btrfs.o] Error 1 (ignored) ... fs/btrfs/btrfs.o: No such file or directory (ignored) Cc: Rusty Russell <rusty@rustcorp.com.au> Cc: Michael Marek <mmarek@suse.cz> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
2013-09-23 05:53:54 +00:00
break;
case 'o':
dump_write = optarg;
break;
case 'a':
all_versions = true;
break;
case 'T':
files_source = optarg;
break;
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
case 't':
trim_unused_exports = true;
break;
case 'u':
unused_exports_white_list = optarg;
break;
case 'W':
extra_warn = true;
break;
case 'w':
warn_unresolved = true;
break;
case 'E':
sec_mismatch_warn_only = false;
break;
case 'N':
allow_missing_ns_imports = true;
break;
case 'd':
missing_namespace_deps = optarg;
break;
default:
exit(1);
}
}
list_for_each_entry_safe(dl, dl2, &dump_lists, list) {
read_dump(dl->file);
list_del(&dl->list);
free(dl);
}
while (optind < argc)
read_symbols(argv[optind++]);
if (files_source)
read_symbols_from_files(files_source);
kbuild: implement CONFIG_TRIM_UNUSED_KSYMS without recursion When CONFIG_TRIM_UNUSED_KSYMS is enabled, Kbuild recursively traverses the directory tree to determine which EXPORT_SYMBOL to trim. If an EXPORT_SYMBOL turns out to be unused by anyone, Kbuild begins the second traverse, where some source files are recompiled with their EXPORT_SYMBOL() tuned into a no-op. Linus stated negative opinions about this slowness in commits: - 5cf0fd591f2e ("Kbuild: disable TRIM_UNUSED_KSYMS option") - a555bdd0c58c ("Kbuild: enable TRIM_UNUSED_KSYMS again, with some guarding") We can do this better now. The final data structures of EXPORT_SYMBOL are generated by the modpost stage, so modpost can selectively emit KSYMTAB entries that are really used by modules. Commit f73edc8951b2 ("kbuild: unify two modpost invocations") is another ground-work to do this in a one-pass algorithm. With the list of modules, modpost sets sym->used if it is used by a module. modpost emits KSYMTAB only for symbols with sym->used==true. BTW, Nicolas explained why the trimming was implemented with recursion: https://lore.kernel.org/all/2o2rpn97-79nq-p7s2-nq5-8p83391473r@syhkavp.arg/ Actually, we never achieved that level of optimization where the chain reaction of trimming comes into play because: - CONFIG_LTO_CLANG cannot remove any unused symbols - CONFIG_LD_DEAD_CODE_DATA_ELIMINATION is enabled only for vmlinux, but not modules If deeper trimming is required, we need to revisit this, but I guess that is unlikely to happen. Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2023-06-11 15:50:57 +00:00
list_for_each_entry(mod, &modules, list) {
if (mod->from_dump || mod->is_vmlinux)
continue;
check_modname_len(mod);
check_exports(mod);
}
if (unused_exports_white_list)
handle_white_list_exports(unused_exports_white_list);
list_for_each_entry(mod, &modules, list) {
if (mod->from_dump)
continue;
2022-05-13 11:39:22 +00:00
if (mod->is_vmlinux)
write_vmlinux_export_c_file(mod);
else
write_mod_c_file(mod);
}
if (missing_namespace_deps)
write_namespace_deps_files(missing_namespace_deps);
if (dump_write)
write_dump(dump_write);
if (sec_mismatch_count && !sec_mismatch_warn_only)
error("Section mismatches detected.\n"
"Set CONFIG_SECTION_MISMATCH_WARN_ONLY=y to allow them.\n");
kbuild: fix false-positive modpost warning when all symbols are trimmed Nathan reports that the mips defconfig emits the following warning: WARNING: modpost: Symbol info of vmlinux is missing. Unresolved symbol check will be entirely skipped. This false-positive happens when CONFIG_TRIM_UNUSED_KSYMS is enabled, but no CONFIG option is set to 'm'. Commit a0590473c5e6 ("nfs: fix PNFS_FLEXFILE_LAYOUT Kconfig default") turned the last 'm' into 'y' for the mips defconfig, and uncovered this issue. In this case, the module feature itself is enabled, but we have no module to build. As a result, CONFIG_TRIM_UNUSED_KSYMS drops all the instances of EXPORT_SYMBOL. Then, modpost wrongly assumes vmlinux is missing because vmlinux.symvers is empty. (As another false-positive case, you can create a module that does not use any symbol of vmlinux). The current behavior is to entirely suppress the unresolved symbol warnings when vmlinux is missing just because there are too many. I found the origin of this code in the historical git tree. [1] If this is a matter of noisiness, I think modpost can display the first 10 warnings, and the number of suppressed warnings at the end. You will get a bit noisier logs when you run 'make modules' without vmlinux, but such warnings are better to show because you never know the resulting modules are actually loadable or not. This commit changes the following: - If any of input *.symver files is missing, pass -w option to let the module build keep going with warnings instead of errors. - If there are too many (10+) unresolved symbol warnings, show only the first 10, and also the number of suppressed warnings. [1]: https://git.kernel.org/pub/scm/linux/kernel/git/history/history.git/commit/?id=1cc0e0529569bf6a94f6d49770aa6d4b599d2c46 Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
2021-03-25 18:54:11 +00:00
if (nr_unresolved > MAX_UNRESOLVED_REPORTS)
warn("suppressed %u unresolved symbol warnings because there were too many)\n",
nr_unresolved - MAX_UNRESOLVED_REPORTS);
return error_occurred ? 1 : 0;
}