Specifically fix the subsection and drop bogus reference to the GNU/Linux.
Reported-by: Patrick Higgins <higgi1pt@gmail.com>
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Commit 3d8439da8 (grub-install: Locale depends on nls) attempted to avoid
copying locale files to the target directory when NLS was disabled.
However the test is inverted, and it does the opposite.
Signed-off-by: Martin Whitaker <fsf@martin-whitaker.me.uk>
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Commit 781b3e5efc (tftp: Do not use priority queue) caused a regression
when fetching files over TFTP whose size is bigger than 65535 * block size.
grub> linux /images/pxeboot/vmlinuz
grub> echo $?
0
grub> initrd /images/pxeboot/initrd.img
error: timeout reading '/images/pxeboot/initrd.img'.
grub> echo $?
28
It is caused by the block number counter being a 16-bit field, which leads
to a maximum file size of ((1 << 16) - 1) * block size. Because GRUB sets
the block size to 1024 octets (by using the TFTP Blocksize Option from RFC
2348 [0]), the maximum file size that can be transferred is 67107840 bytes.
The TFTP PROTOCOL (REVISION 2) RFC 1350 [1] does not mention what a client
should do when a file size is bigger than the maximum, but most TFTP hosts
support the block number counter to be rolled over. That is, acking a data
packet with a block number of 0 is taken as if the 65356th block was acked.
It was working before because the block counter roll-over was happening due
an overflow. But that got fixed by the mentioned commit, which led to the
regression when attempting to fetch files larger than the maximum size.
To allow TFTP file transfers of unlimited size again, re-introduce a block
counter roll-over so the data packets are acked preventing the timeouts.
[0]: https://tools.ietf.org/html/rfc2348
[1]: https://tools.ietf.org/html/rfc1350
Fixes: 781b3e5efc (tftp: Do not use priority queue)
Suggested-by: Peter Jones <pjones@redhat.com>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Here dev is a grub_cryptodisk_t and dev->offset is offset in sectors of size
native to the cryptodisk device. The sector is correctly transformed into
native grub sector size, but then added to dev->offset which is not
transformed. It would be nice if the type system would help us with this.
Signed-off-by: Glenn Washburn <development@efficientek.com>
Reviewed-by: Patrick Steinhardt <ps@pks.im>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
While we already set up error messages in both luks2_verify_key() and
luks2_decrypt_key(), we do not ever print them. This makes it really
hard to discover why a given key actually failed to decrypt a disk.
Improve this by including the error message in the user-visible output.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
When configuring a LUKS disk, we copy over the UUID from the LUKS header
into the new grub_cryptodisk_t structure via grub_memcpy(). As size
we mistakenly use the size of the grub_cryptodisk_t UUID field, which
is guaranteed to be strictly bigger than the LUKS UUID field we're
copying. As a result, the copy always goes out-of-bounds and copies some
garbage from other surrounding fields. During runtime, this isn't
noticed due to the fact that we always NUL-terminate the UUID and thus
never hit the trailing garbage.
Fix the issue by using the size of the local stripped UUID field.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The C standard does not allow for typedef redefinitions, even if they
map to the same underlying type. In order to avoid including the
jsmn.h in json.h and thus exposing jsmn's internals, we have exactly
such a forward-declaring typedef in json.h. If enforcing the GNU99 C
standard, clang may generate a warning about this non-standard
construct.
Fix the issue by using a simple "struct jsmntok" forward declaration
instead of using a typedef.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Tested-by: Chuck Tuffli <chuck@freebsd.org>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
These could be triggered by a crafted filesystem with very large files.
Fixes: CVE-2020-15707
Signed-off-by: Colin Watson <cjwatson@debian.org>
Reviewed-by: Jan Setje-Eilers <jan.setjeeilers@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
commit 92bfc33db9 ("efi: Free malloc regions on exit")
introduced memory freeing in grub_efi_fini(), which is
used not only by exit path but by halt/reboot one as well.
As result of memory freeing, code and data regions used by
modules, such as halt, reboot, acpi (used by halt) also got
freed. After return to module code, CPU executes, filled
by UEFI firmware (tested with edk2), 0xAFAFAFAF pattern as
a code. Which leads to #UD exception later.
grub> halt
!!!! X64 Exception Type - 06(#UD - Invalid Opcode) CPU Apic ID - 00000000 !!!!
RIP - 0000000003F4EC28, CS - 0000000000000038, RFLAGS - 0000000000200246
RAX - 0000000000000000, RCX - 00000000061DA188, RDX - 0A74C0854DC35D41
RBX - 0000000003E10E08, RSP - 0000000007F0F860, RBP - 0000000000000000
RSI - 00000000064DB768, RDI - 000000000832C5C3
R8 - 0000000000000002, R9 - 0000000000000000, R10 - 00000000061E2E52
R11 - 0000000000000020, R12 - 0000000003EE5C1F, R13 - 00000000061E0FF4
R14 - 0000000003E10D80, R15 - 00000000061E2F60
DS - 0000000000000030, ES - 0000000000000030, FS - 0000000000000030
GS - 0000000000000030, SS - 0000000000000030
CR0 - 0000000080010033, CR2 - 0000000000000000, CR3 - 0000000007C01000
CR4 - 0000000000000668, CR8 - 0000000000000000
DR0 - 0000000000000000, DR1 - 0000000000000000, DR2 - 0000000000000000
DR3 - 0000000000000000, DR6 - 00000000FFFF0FF0, DR7 - 0000000000000400
GDTR - 00000000079EEA98 0000000000000047, LDTR - 0000000000000000
IDTR - 0000000007598018 0000000000000FFF, TR - 0000000000000000
FXSAVE_STATE - 0000000007F0F4C0
Proposal here is to continue to free allocated memory for
exit boot services path but keep it for halt/reboot path
as it won't be much security concern here.
Introduced GRUB_LOADER_FLAG_EFI_KEEP_ALLOCATED_MEMORY
loader flag to be used by efi halt/reboot path.
Signed-off-by: Alexey Makhalov <amakhalov@vmware.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Without any error propagated to the caller, make_file_path()
would then try to advance the invalid device path node with
GRUB_EFI_NEXT_DEVICE_PATH(), which would fail, returning a NULL
pointer that would subsequently be dereferenced. Hence, propagate
errors from copy_file_path().
Signed-off-by: Chris Coulson <chris.coulson@canonical.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Several places we take the length of a device path and subtract 4 from
it, without ever checking that it's >= 4. There are also cases where
this kind of malformation will result in unpredictable iteration,
including treating the length from one dp node as the type in the next
node. These are all errors, no matter where the data comes from.
This patch adds a checking macro, GRUB_EFI_DEVICE_PATH_VALID(), which
can be used in several places, and makes GRUB_EFI_NEXT_DEVICE_PATH()
return NULL and GRUB_EFI_END_ENTIRE_DEVICE_PATH() evaluate as true when
the length is too small. Additionally, it makes several places in the
code check for and return errors in these cases.
Signed-off-by: Peter Jones <pjones@redhat.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The grub_free() implementation in grub-core/kern/mm.c safely handles
NULL pointers, and code at many places depends on this. We don't know
that the same is true on all host OSes, so we need to handle the same
behavior in grub-emu's implementation.
Signed-off-by: Peter Jones <pjones@redhat.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
It appears to be possible to make a (possibly invalid) lvm PV with
a metadata size field that overflows our type when adding it to the
address we've allocated. Even if it doesn't, it may be possible to do so
with the math using the outcome of that as an operand. Check them both.
Signed-off-by: Peter Jones <pjones@redhat.com>
Signed-off-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Both node->size and node->namelen come from the supplied filesystem,
which may be user-supplied. We can't trust them for the math unless we
know they don't overflow. Making sure they go through grub_add() or
grub_calloc() first will give us that.
Signed-off-by: Peter Jones <pjones@redhat.com>
Reviewed-by: Darren Kenny <darren.kenny@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Current implementation of grub_relocator_alloc_chunk_align()
does not allow allocation of the top byte.
Assuming input args are:
max_addr = 0xfffff000;
size = 0x1000;
And this is valid. But following overflow protection will
unnecessarily move max_addr one byte down (to 0xffffefff):
if (max_addr > ~size)
max_addr = ~size;
~size + 1 will fix the situation. In addition, check size
for non zero to do not zero max_addr.
Signed-off-by: Alexey Makhalov <amakhalov@vmware.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Defining a new function with the same name as a previously defined
function causes the grub_script and associated resources for the
previous function to be freed. If the previous function is currently
executing when a function with the same name is defined, this results
in use-after-frees when processing subsequent commands in the original
function.
Instead, reject a new function definition if it has the same name as
a previously defined function, and that function is currently being
executed. Although a behavioural change, this should be backwards
compatible with existing configurations because they can't be
dependent on the current behaviour without being broken.
Fixes: CVE-2020-15706
Signed-off-by: Chris Coulson <chris.coulson@canonical.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
This commit introduces integer underflow mitigation in max_addr calculation
in grub_relocator_alloc_chunk_align() invocation.
It consists of 2 fixes:
1. Introduced grub_relocator_alloc_chunk_align_safe() wrapper function to perform
sanity check for min/max and size values, and to make safe invocation of
grub_relocator_alloc_chunk_align() with validated max_addr value. Replace all
invocations such as grub_relocator_alloc_chunk_align(..., min_addr, max_addr - size, size, ...)
by grub_relocator_alloc_chunk_align_safe(..., min_addr, max_addr, size, ...).
2. Introduced UP_TO_TOP32(s) macro for the cases where max_addr is 32-bit top
address (0xffffffff - size + 1) or similar.
Signed-off-by: Alexey Makhalov <amakhalov@vmware.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Use arithmetic macros from safemath.h to accomplish it. In this commit,
I didn't want to be too paranoid to check every possible math equation
for overflow/underflow. Only obvious places (with non zero chance of
overflow/underflow) were refactored.
Signed-off-by: Alexey Makhalov <amakhalov@vmware.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
There is not need to reassemble the order of blocks. Per RFC 1350,
server must wait for the ACK, before sending next block. Data packets
can be served immediately without putting them to priority queue.
Logic to handle incoming packet is this:
- if packet block id equal to expected block id, then
process the packet,
- if packet block id is less than expected - this is retransmit
of old packet, then ACK it and drop the packet,
- if packet block id is more than expected - that shouldn't
happen, just drop the packet.
It makes the tftp receive path code simpler, smaller and faster.
As a benefit, this change fixes CID# 73624 and CID# 96690, caused
by following while loop:
while (cmp_block (grub_be_to_cpu16 (tftph->u.data.block), data->block + 1) == 0)
where tftph pointer is not moving from one iteration to another, causing
to serve same packet again. Luckily, double serving didn't happen due to
data->block++ during the first iteration.
Fixes: CID 73624, CID 96690
Signed-off-by: Alexey Makhalov <amakhalov@vmware.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
This requires a very weird input from the serial interface but can cause
an overflow in input_buf (keys) overwriting the next variable (npending)
with the user choice:
(pahole output)
struct grub_terminfo_input_state {
int input_buf[6]; /* 0 24 */
int npending; /* 24 4 */ <- CORRUPT
...snip...
The magic string requires causing this is "ESC,O,],0,1,2,q" and we overflow
npending with "q" (aka increase npending to 161). The simplest fix is to
just to disallow overwrites input_buf, which exactly what this patch does.
Fixes: CID 292449
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The two dimensional array p->posSlotEncoder[4][64] is being dereferenced
using the GetLenToPosState() macro which checks if len is less than 5,
and if so subtracts 2 from it. If len = 0, that is 0 - 2 = 4294967294.
Obviously we don't want to dereference that far out so we check if the
position found is greater or equal kNumLenToPosStates (4) and bail out.
N.B.: Upstream LZMA 18.05 and later has this function completely rewritten
without any history.
Fixes: CID 51526
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
When grub_json_parse() succeeds, it returns the root object which
contains a pointer to the provided JSON string. Callers are
responsible for ensuring that this string outlives the root
object and for freeing its memory when it's no longer needed.
If grub_json_parse() fails to parse the provided JSON string,
it frees the string before returning an error. This results
in a double free in luks2_recover_key(), which also frees the
same string after grub_json_parse() returns an error.
This changes grub_json_parse() to never free the JSON string
passed to it, and updates the documentation for it to make it
clear that callers are responsible for ensuring that the string
outlives the root JSON object.
Fixes: CID 292465
Signed-off-by: Chris Coulson <chris.coulson@canonical.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
grub_xnu_devprop_add_property() should not free utf8 and utf16 as it get
allocated and freed in the caller.
Minor improvement: do prop fields initialization after memory allocations.
Fixes: CID 292442, CID 292457, CID 292460, CID 292466
Signed-off-by: Alexey Makhalov <amakhalov@vmware.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
self->bitmap should be zeroed after free. Otherwise, there is a chance
to double free (USE_AFTER_FREE) it later in rescale_image().
Fixes: CID 292472
Signed-off-by: Alexey Makhalov <amakhalov@vmware.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The GRUB font file can have one NAME section only. Though if somebody
crafts a broken font file with many NAME sections and loads it then the
GRUB leaks memory. So, prevent against that by loading first NAME
section and failing in controlled way on following one.
Reported-by: Chris Coulson <chris.coulson@canonical.com>
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Reviewed-by: Jan Setje-Eilers <jan.setjeeilers@oracle.com>
This attempts to fix the places where we do the following where
arithmetic_expr may include unvalidated data:
X = grub_malloc(arithmetic_expr);
It accomplishes this by doing the arithmetic ahead of time using grub_add(),
grub_sub(), grub_mul() and testing for overflow before proceeding.
Among other issues, this fixes:
- allocation of integer overflow in grub_video_bitmap_create()
reported by Chris Coulson,
- allocation of integer overflow in grub_png_decode_image_header()
reported by Chris Coulson,
- allocation of integer overflow in grub_squash_read_symlink()
reported by Chris Coulson,
- allocation of integer overflow in grub_ext2_read_symlink()
reported by Chris Coulson,
- allocation of integer overflow in read_section_as_string()
reported by Chris Coulson.
Fixes: CVE-2020-14309, CVE-2020-14310, CVE-2020-14311
Signed-off-by: Peter Jones <pjones@redhat.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
This modifies most of the places we do some form of:
X = malloc(Y * Z);
to use calloc(Y, Z) instead.
Among other issues, this fixes:
- allocation of integer overflow in grub_png_decode_image_header()
reported by Chris Coulson,
- allocation of integer overflow in luks_recover_key()
reported by Chris Coulson,
- allocation of integer overflow in grub_lvm_detect()
reported by Chris Coulson.
Fixes: CVE-2020-14308
Signed-off-by: Peter Jones <pjones@redhat.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
This tries to make sure that everywhere in this source tree, we always have
an appropriate version of calloc() (i.e. grub_calloc(), xcalloc(), etc.)
available, and that they all safely check for overflow and return NULL when
it would occur.
Signed-off-by: Peter Jones <pjones@redhat.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
This adds a new header, include/grub/safemath.h, that includes easy to
use wrappers for __builtin_{add,sub,mul}_overflow() declared like:
bool OP(a, b, res)
where OP is grub_add, grub_sub or grub_mul. OP() returns true in the
case where the operation would overflow and res is not modified.
Otherwise, false is returned and the operation is executed.
These arithmetic primitives require newer compiler versions. So, bump
these requirements in the INSTALL file too.
Signed-off-by: Peter Jones <pjones@redhat.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
When presented with a command that can't be tokenized to anything
smaller than YYLMAX characters, the parser calls YY_FATAL_ERROR(errmsg),
expecting that will stop further processing, as such:
#define YY_DO_BEFORE_ACTION \
yyg->yytext_ptr = yy_bp; \
yyleng = (int) (yy_cp - yy_bp); \
yyg->yy_hold_char = *yy_cp; \
*yy_cp = '\0'; \
if ( yyleng >= YYLMAX ) \
YY_FATAL_ERROR( "token too large, exceeds YYLMAX" ); \
yy_flex_strncpy( yytext, yyg->yytext_ptr, yyleng + 1 , yyscanner); \
yyg->yy_c_buf_p = yy_cp;
The code flex generates expects that YY_FATAL_ERROR() will either return
for it or do some form of longjmp(), or handle the error in some way at
least, and so the strncpy() call isn't in an "else" clause, and thus if
YY_FATAL_ERROR() is *not* actually fatal, it does the call with the
questionable limit, and predictable results ensue.
Unfortunately, our implementation of YY_FATAL_ERROR() is:
#define YY_FATAL_ERROR(msg) \
do { \
grub_printf (_("fatal error: %s\n"), _(msg)); \
} while (0)
The same pattern exists in yyless(), and similar problems exist in users
of YY_INPUT(), several places in the main parsing loop,
yy_get_next_buffer(), yy_load_buffer_state(), yyensure_buffer_stack,
yy_scan_buffer(), etc.
All of these callers expect YY_FATAL_ERROR() to actually be fatal, and
the things they do if it returns after calling it are wildly unsafe.
Fixes: CVE-2020-10713
Signed-off-by: Peter Jones <pjones@redhat.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
When booting on an ARMv8 core that implements either CTR.IDC or CTR.DIC
(indicating that some of the cache maintenance operations can be
removed when dealing with I/D-cache coherency, GRUB dies with a
"Unsupported cache type 0x........" message.
This is pretty likely to happen when running in a virtual machine
hosted on an arm64 machine (I've triggered it on a system built around
a bunch of Cortex-A55 cores, which implements CTR.IDC).
It turns out that the way GRUB deals with the CTR register is a bit
harsh for anything from ARMv7 onwards. The layout of the register is
backward compatible, meaning that nothing that gets added is allowed to
break earlier behaviour. In this case, ignoring IDC is completely fine,
and only results in unnecessary cache maintenance.
We can thus avoid being paranoid, and align the 32bit behaviour with
its 64bit equivalent.
This patch has the added benefit that it gets rid of a (gnu-specific)
case range too.
Signed-off-by: Marc Zyngier <maz@kernel.org>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
XSM is enabled by adding "flask=enforcing" as a Xen command line
argument, and providing the policy file as a grub module.
We make entries for both with and without XSM. If XSM is not compiled
into Xen, then there are no policy files, so no change to the boot
options.
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
file_is_not_sym() currently only checks for xen-syms. Extend it to
disregard xenpolicy (XSM policy files) and files ending .config (which
are built by the Xen upstream build system in some configurations and
can therefore end up in /boot).
Rename the function accordingly, to file_is_not_xen_garbage().
Signed-off-by: Ian Jackson <ian.jackson@eu.citrix.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
Nested functions are not supported in C, but are permitted as an extension
in the GNU C dialect. Commit cb2f15c544 ("normal/main: Search for specific
config files for netboot") added a nested function which caused the build
to break when compiling with clang.
Break that out into a static helper function to make the code portable again.
Reported-by: Daniel Axtens <dja@axtens.net>
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Tested-by: Daniel Axtens <dja@axtens.net>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
The module is only enabled for x86_64, but there's nothing specific to
x86_64 in the implementation and can be enabled for all EFI platforms.
Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
..to reflect the GRUB build reality in them.
Additionally, fix text formatting a bit.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
Commit d5a32255d (misc: Make grub_strtol() "end" pointers have safer
const qualifiers) introduced "restrict" keyword into some functions
definitions. This keyword was introduced in C99 standard. However, some
compilers by default may use C89 or something different. This behavior
leads to the breakage during builds when c89 or gnu89 is in force. So,
let's set gnu99 C language standard for all compilers by default. This
way a bit random build issue will be fixed and the GRUB source will be
build consistently regardless of type and version of the compiler.
It was decided to use gnu99 C language standard because it fixes the
issue mentioned above and also provides some useful extensions which are
used here and there in the GRUB source. Potentially we can use gnu11
too. However, this may reduce pool of older compilers which can be used
to build the GRUB. So, let's live with gnu99 until we discover that we
strongly require a feature from newer C standard.
The user is still able to override C language standard using relevant
*_CFLAGS variables.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
grub_tpm_log_event() and grub_tpm_measure() are two functions that
have the same effect. So, keep grub_tpm_log_event() and rename it
to grub_tpm_measure(). This way we get also a more clear semantics.
Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
..because -iname cannot be used to match paths.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Leif Lindholm <leif@nuviainc.com>
Reviewed-by: Daniel Axtens <dja@axtens.net>
