Re: [PATCH 5.4 18/43] arm64: entry: Add macro for reading symbol addresses from the trampoline

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 





On 3/18/2022 5:11 AM, James Morse wrote:
Hi Florian,

On 3/17/22 8:48 PM, Florian Fainelli wrote:
On 3/17/22 5:45 AM, Greg Kroah-Hartman wrote:
From: James Morse <james.morse@xxxxxxx>

commit b28a8eebe81c186fdb1a0078263b30576c8e1f42 upstream.

The trampoline code needs to use the address of symbols in the wider
kernel, e.g. vectors. PC-relative addressing wouldn't work as the
trampoline code doesn't run at the address the linker expected.

tramp_ventry uses a literal pool, unless CONFIG_RANDOMIZE_BASE is
set, in which case it uses the data page as a literal pool because
the data page can be unmapped when running in user-space, which is
required for CPUs vulnerable to meltdown.

Pull this logic out as a macro, instead of adding a third copy
of it.

Reviewed-by: Catalin Marinas <catalin.marinas@xxxxxxx>
Signed-off-by: James Morse <james.morse@xxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>

This commit causes a linking failure with CONFIG_ARM_SDE_INTERFACE=y
enabled in the kernel:

   LD      .tmp_vmlinux.kallsyms1
/local/users/fainelli/buildroot/output/arm64/host/bin/aarch64-linux-ld:
arch/arm64/kernel/entry.o: in function `__sdei_asm_exit_trampoline':
/local/users/fainelli/buildroot/output/arm64/build/linux-custom/arch/arm64/kernel/entry.S:1352:
undefined reference to `__sdei_asm_trampoline_next_handler'
make[2]: *** [Makefile:1100: vmlinux] Error 1
make[1]: *** [package/pkg-generic.mk:295:
/local/users/fainelli/buildroot/output/arm64/build/linux-custom/.stamp_built]
Error 2
make: *** [Makefile:27: _all] Error 2

... and with CONFIG_RANDOMIZE_BASE turned off, which is why allyesconfig didn't catch it.

Yes that is correct CONFIG_RANDOMIZE_BASE is turned off in the configuration file I used.

This is because I kept the next_handler bit of the label when it conflicted, which isn't needed
because the __entry_tramp bit added by the macro serves the same purpose.

The below diff fixes it:
----------%<----------
diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S
index e4b5a15c2e2e..cfc0bb6c49f7 100644
--- a/arch/arm64/kernel/entry.S
+++ b/arch/arm64/kernel/entry.S
@@ -1190,7 +1190,7 @@ __entry_tramp_data_start:
  __entry_tramp_data_vectors:
         .quad   vectors
  #ifdef CONFIG_ARM_SDE_INTERFACE
-__entry_tramp_data___sdei_asm_trampoline_next_handler:
+__entry_tramp_data___sdei_asm_handler:
         .quad   __sdei_asm_handler
  #endif /* CONFIG_ARM_SDE_INTERFACE */
         .popsection                             // .rodata
@@ -1319,7 +1319,7 @@ ENTRY(__sdei_asm_entry_trampoline)
          */
  1:     str     x4, [x1, #(SDEI_EVENT_INTREGS + S_ORIG_ADDR_LIMIT)]

-       tramp_data_read_var     x4, __sdei_asm_trampoline_next_handler
+       tramp_data_read_var     x4, __sdei_asm_handler
         br      x4
  ENDPROC(__sdei_asm_entry_trampoline)
  NOKPROBE(__sdei_asm_entry_trampoline)
----------%<----------

Good news - this didn't happen with v5.10.

I don't see this in v5.4.185 yet.

Greg/Sasha, what is least work for you?:
A new version of this patch,
A fixup on top of the series,
Reposting the series with this fixed.

FWIW, applying your patch on top of the entire 5.4.186-rc1 series will fail with some rejects so that would require you to fixup the patch instead, and then I suppose the whole -rc1 now becomes an -rc2?



Thanks for catching this!

Thanks for the quick turnaround:

Tested-by: Florian Fainelli <f.fainelli@xxxxxxxxx>
--
Florian



[Index of Archives]     [Linux Kernel]     [Kernel Development Newbies]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite Hiking]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux