linux-stable/arch
Nicholas Piggin 0138ba5783 powerpc/64/signal: Balance return predictor stack in signal trampoline
Returning from an interrupt or syscall to a signal handler currently
begins execution directly at the handler's entry point, with LR set to
the address of the sigreturn trampoline. When the signal handler
function returns, it runs the trampoline. It looks like this:

    # interrupt at user address xyz
    # kernel stuff... signal is raised
    rfid
    # void handler(int sig)
    addis 2,12,.TOC.-.LCF0@ha
    addi 2,2,.TOC.-.LCF0@l
    mflr 0
    std 0,16(1)
    stdu 1,-96(1)
    # handler stuff
    ld 0,16(1)
    mtlr 0
    blr
    # __kernel_sigtramp_rt64
    addi    r1,r1,__SIGNAL_FRAMESIZE
    li      r0,__NR_rt_sigreturn
    sc
    # kernel executes rt_sigreturn
    rfid
    # back to user address xyz

Note the blr with no matching bl. This can corrupt the return
predictor.

Solve this by instead resuming execution at the signal trampoline
which then calls the signal handler. qtrace-tools link_stack checker
confirms the entire user/kernel/vdso cycle is balanced after this
patch, whereas it's not upstream.

Alan confirms the dwarf unwind info still looks good. gdb still
recognises the signal frame and can step into parent frames if it
break inside a signal handler.

Performance is pretty noisy, not a very significant change on a POWER9
here, but branch misses are consistently a lot lower on a
microbenchmark:

 Performance counter stats for './signal':

       13,085.72 msec task-clock                #    1.000 CPUs utilized
  45,024,760,101      cycles                    #    3.441 GHz
  65,102,895,542      instructions              #    1.45  insn per cycle
  11,271,673,787      branches                  #  861.372 M/sec
      59,468,979      branch-misses             #    0.53% of all branches

       12,989.09 msec task-clock                #    1.000 CPUs utilized
  44,692,719,559      cycles                    #    3.441 GHz
  65,109,984,964      instructions              #    1.46  insn per cycle
  11,282,136,057      branches                  #  868.585 M/sec
      39,786,942      branch-misses             #    0.35% of all branches

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200511101952.1463138-1-npiggin@gmail.com
2020-07-15 11:08:27 +10:00
..
alpha Kbuild updates for v5.8 (2nd) 2020-06-13 13:29:16 -07:00
arc treewide: replace '---help---' in Kconfig files with 'help' 2020-06-14 01:57:21 +09:00
arm maccess: make get_kernel_nofault() check for minimal type compatibility 2020-06-18 12:10:37 -07:00
arm64 Kbuild fixes for v5.8 2020-06-21 12:44:52 -07:00
c6x This time around we have 4 lines of diff in the core framework, removing a 2020-06-10 11:42:19 -07:00
csky maccess: rename probe_kernel_{read,write} to copy_{from,to}_kernel_nofault 2020-06-17 10:57:41 -07:00
h8300 This time around we have 4 lines of diff in the core framework, removing a 2020-06-10 11:42:19 -07:00
hexagon treewide: replace '---help---' in Kconfig files with 'help' 2020-06-14 01:57:21 +09:00
ia64 Merge branch 'hch' (maccess patches from Christoph Hellwig) 2020-06-18 12:35:51 -07:00
m68k Kbuild updates for v5.8 (2nd) 2020-06-13 13:29:16 -07:00
microblaze mmap locking API: convert mmap_sem comments 2020-06-09 09:39:14 -07:00
mips maccess: rename probe_kernel_{read,write} to copy_{from,to}_kernel_nofault 2020-06-17 10:57:41 -07:00
nds32 maccess: rename probe_kernel_{read,write} to copy_{from,to}_kernel_nofault 2020-06-17 10:57:41 -07:00
nios2 nios2 update for v5.8-rc1 2020-06-12 11:55:11 -07:00
openrisc OpenRISC updates for 5.8 2020-06-13 10:54:09 -07:00
parisc maccess: make get_kernel_nofault() check for minimal type compatibility 2020-06-18 12:10:37 -07:00
powerpc powerpc/64/signal: Balance return predictor stack in signal trampoline 2020-07-15 11:08:27 +10:00
riscv RISC-V Fixes for 5.8-rc2 2020-06-20 12:14:29 -07:00
s390 s390 fixes for 5.8-rc2 2020-06-20 12:31:08 -07:00
sh maccess: rename probe_kernel_address to get_kernel_nofault 2020-06-18 11:14:40 -07:00
sparc treewide: replace '---help---' in Kconfig files with 'help' 2020-06-14 01:57:21 +09:00
um maccess: rename probe_kernel_{read,write} to copy_{from,to}_kernel_nofault 2020-06-17 10:57:41 -07:00
unicore32 This time around we have 4 lines of diff in the core framework, removing a 2020-06-10 11:42:19 -07:00
x86 Merge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux 2020-06-20 19:18:27 -07:00
xtensa mmap locking API: convert mmap_sem API comments 2020-06-09 09:39:14 -07:00
.gitignore
Kconfig treewide: replace '---help---' in Kconfig files with 'help' 2020-06-14 01:57:21 +09:00