diff mbox

[v3,2/3] powerpc/reloc64: add support for 32-bit CRC pseudo-symbols

Message ID 1477585631-18574-3-git-send-email-ard.biesheuvel@linaro.org
State New
Headers show

Commit Message

Ard Biesheuvel Oct. 27, 2016, 4:27 p.m. UTC 
In preparation of modifying the core modversions code to emit the CRCs
as 32-bit quantities, ensure that 64-bit PowerPC will be able to deal
with this when CONFIG_RELOCATABLE=y, in which case the CRCs will be
emitted into the final ELF binary as R_PPC64_ADDR32 relocations.

Since 32-bit relocations cannot be used to relocate memory addresses on
64-bit architectures, and since the CRC pseudo-symbol references are
emitted as anonymous relocations (i.e., against the NULL symbol in the
.dynsym section) with the final value recorded in the addend (*), we
can disregard any relocations where the symbol index != 0.

* Note that unsatisfied CRC pseudo-symbol references are emitted as
  R_PPC64_ADDR32 relocations against named symbols that are typed as
  weak undefined in the .dynsym symbol table. These can simply be
  ignored (as before), considering that zero CRCs are interpreted as
  missing, and the module code deals with that accordingly.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>

---
 arch/powerpc/kernel/reloc_64.S | 22 ++++++++++++++++----
 arch/powerpc/relocs_check.sh   |  3 ++-
 2 files changed, 20 insertions(+), 5 deletions(-)

-- 
2.7.4

Comments

Michael Ellerman Nov. 25, 2016, 11:29 a.m. UTC | #1 
Ard Biesheuvel <ard.biesheuvel@linaro.org> writes:

> diff --git a/arch/powerpc/relocs_check.sh b/arch/powerpc/relocs_check.sh

> index ec2d5c835170..2f510fbc87da 100755

> --- a/arch/powerpc/relocs_check.sh

> +++ b/arch/powerpc/relocs_check.sh

> @@ -43,7 +43,8 @@ R_PPC_ADDR16_HA

>  R_PPC_RELATIVE

>  R_PPC_NONE' |

>  	grep -E -v '\<R_PPC64_ADDR64[[:space:]]+mach_' |

> -	grep -E -v '\<R_PPC64_ADDR64[[:space:]]+__crc_'

> +	grep -E -v '\<R_PPC64_ADDR64[[:space:]]+__crc_' |

> +	grep -E -v '\<R_PPC64_ADDR32[[:space:]]+\*ABS\*'


I'm still getting:

WARNING: 24 bad relocations
c000000000d307c4 R_PPC64_ADDR32    __crc___arch_hweight16
c000000000d307c8 R_PPC64_ADDR32    __crc___arch_hweight32
c000000000d307cc R_PPC64_ADDR32    __crc___arch_hweight64
c000000000d307d0 R_PPC64_ADDR32    __crc___arch_hweight8
c000000000d30848 R_PPC64_ADDR32    __crc___bswapdi2
c000000000d30854 R_PPC64_ADDR32    __crc___clear_user
c000000000d30868 R_PPC64_ADDR32    __crc___copy_tofrom_user
c000000000d30d4c R_PPC64_ADDR32    __crc__mcount
c000000000d31344 R_PPC64_ADDR32    __crc_copy_page
c000000000d3141c R_PPC64_ADDR32    __crc_current_stack_pointer
c000000000d31840 R_PPC64_ADDR32    __crc_empty_zero_page
c000000000d31a7c R_PPC64_ADDR32    __crc_flush_dcache_range
c000000000d31a84 R_PPC64_ADDR32    __crc_flush_icache_range
c000000000d32608 R_PPC64_ADDR32    __crc_load_fp_state
c000000000d32614 R_PPC64_ADDR32    __crc_load_vr_state
c000000000d32828 R_PPC64_ADDR32    __crc_memchr
c000000000d32830 R_PPC64_ADDR32    __crc_memcmp
c000000000d32834 R_PPC64_ADDR32    __crc_memcpy
c000000000d32840 R_PPC64_ADDR32    __crc_memmove
c000000000d32888 R_PPC64_ADDR32    __crc_memset
c000000000d33c9c R_PPC64_ADDR32    __crc_store_fp_state
c000000000d33ca0 R_PPC64_ADDR32    __crc_store_vr_state
c000000000d33cf0 R_PPC64_ADDR32    __crc_strncmp
c000000000d33cf4 R_PPC64_ADDR32    __crc_strncpy


If I just add those to the whitelist it builds, but then things aren't
happy at boot:


[    7.607687] kvm: disagrees about version of symbol module_layout
[    7.846799] virtio: disagrees about version of symbol module_layout
[   22.012615] crc32c_vpmsum: disagrees about version of symbol module_layout
[   22.012959] libcrc32c: disagrees about version of symbol module_layout


cheers
Michael Ellerman Dec. 1, 2016, 9:39 a.m. UTC | #2 
Ard Biesheuvel <ard.biesheuvel@linaro.org> writes:
> On 25 November 2016 at 11:29, Michael Ellerman <mpe@ellerman.id.au> wrote:

>> Ard Biesheuvel <ard.biesheuvel@linaro.org> writes:

>>

>> [    7.607687] kvm: disagrees about version of symbol module_layout

>> [    7.846799] virtio: disagrees about version of symbol module_layout

>> [   22.012615] crc32c_vpmsum: disagrees about version of symbol module_layout

>> [   22.012959] libcrc32c: disagrees about version of symbol module_layout

>>

>

> Sigh. I suppose your modversions fixes are queued for v4.10? It's

> probably best to revisit this after the v4.10 merge window closes

> then, just to make sure I'm not aiming for a moving target.


Actually they were merged into 4.9-rc7 ish.

But I'm still seeing the same as above with this series rebased on top
of that, and I'm a bit short on time to debug it ATM.

So during the 4.10 cycle is probably the best we can hope for, sorry.

cheers
Ard Biesheuvel Dec. 1, 2016, 9:45 a.m. UTC | #3 
On 1 December 2016 at 09:39, Michael Ellerman <mpe@ellerman.id.au> wrote:
> Ard Biesheuvel <ard.biesheuvel@linaro.org> writes:

>> On 25 November 2016 at 11:29, Michael Ellerman <mpe@ellerman.id.au> wrote:

>>> Ard Biesheuvel <ard.biesheuvel@linaro.org> writes:

>>>

>>> [    7.607687] kvm: disagrees about version of symbol module_layout

>>> [    7.846799] virtio: disagrees about version of symbol module_layout

>>> [   22.012615] crc32c_vpmsum: disagrees about version of symbol module_layout

>>> [   22.012959] libcrc32c: disagrees about version of symbol module_layout

>>>

>>

>> Sigh. I suppose your modversions fixes are queued for v4.10? It's

>> probably best to revisit this after the v4.10 merge window closes

>> then, just to make sure I'm not aiming for a moving target.

>

> Actually they were merged into 4.9-rc7 ish.

>

> But I'm still seeing the same as above with this series rebased on top

> of that, and I'm a bit short on time to debug it ATM.

>

> So during the 4.10 cycle is probably the best we can hope for, sorry.

>


Not a problem. The only question is whether 1/3 of this series fixes
an actual bug or not, given that the CONFIG_RELOCATABLE workaround has
been made ppc64 only.

But for the remaining patches, I'm happy to respin after the v4.10
merge window closes, and get something queued for v4.11

-- 
Ard.
Ard Biesheuvel Dec. 1, 2016, 4:28 p.m. UTC | #4 
On 1 December 2016 at 09:45, Ard Biesheuvel <ard.biesheuvel@linaro.org> wrote:
> On 1 December 2016 at 09:39, Michael Ellerman <mpe@ellerman.id.au> wrote:

>> Ard Biesheuvel <ard.biesheuvel@linaro.org> writes:

>>> On 25 November 2016 at 11:29, Michael Ellerman <mpe@ellerman.id.au> wrote:

>>>> Ard Biesheuvel <ard.biesheuvel@linaro.org> writes:

>>>>

>>>> [    7.607687] kvm: disagrees about version of symbol module_layout

>>>> [    7.846799] virtio: disagrees about version of symbol module_layout

>>>> [   22.012615] crc32c_vpmsum: disagrees about version of symbol module_layout

>>>> [   22.012959] libcrc32c: disagrees about version of symbol module_layout

>>>>

>>>

>>> Sigh. I suppose your modversions fixes are queued for v4.10? It's

>>> probably best to revisit this after the v4.10 merge window closes

>>> then, just to make sure I'm not aiming for a moving target.

>>

>> Actually they were merged into 4.9-rc7 ish.

>>

>> But I'm still seeing the same as above with this series rebased on top

>> of that, and I'm a bit short on time to debug it ATM.

>>

>> So during the 4.10 cycle is probably the best we can hope for, sorry.

>>

>

> Not a problem. The only question is whether 1/3 of this series fixes

> an actual bug or not, given that the CONFIG_RELOCATABLE workaround has

> been made ppc64 only.

>

> But for the remaining patches, I'm happy to respin after the v4.10

> merge window closes, and get something queued for v4.11

>


Actually, given the uncertain fate of modversions in general, we may
no longer have to bother by the time the v4.11 merge window opens ...
diff mbox

Patch

diff --git a/arch/powerpc/kernel/reloc_64.S b/arch/powerpc/kernel/reloc_64.S
index d88736fbece6..7927e00be746 100644
--- a/arch/powerpc/kernel/reloc_64.S
+++ b/arch/powerpc/kernel/reloc_64.S
@@ -14,6 +14,7 @@ 
 RELA = 7
 RELACOUNT = 0x6ffffff9
 R_PPC64_RELATIVE = 22
+R_PPC64_ADDR32 = 1
 
 /*
  * r3 = desired final address of kernel
@@ -66,10 +67,10 @@  _GLOBAL(relocate)
 
 	/*
 	 * Run through the list of relocations and process the
-	 * R_PPC64_RELATIVE ones.
+	 * R_PPC64_RELATIVE and R_PPC64_ADDR32 ones.
 	 */
 	mtctr	r8
-5:	ld	r0,8(9)		/* ELF64_R_TYPE(reloc->r_info) */
+5:	ld	r0,8(9)		/* reloc->r_info (type *and* symbol index) */
 	cmpdi	r0,R_PPC64_RELATIVE
 	bne	6f
 	ld	r6,0(r9)	/* reloc->r_offset */
@@ -77,9 +78,22 @@  _GLOBAL(relocate)
 	add	r0,r0,r3
 	stdx	r0,r7,r6
 	addi	r9,r9,24
-	bdnz	5b
+	b	7f
+
+	/*
+	 * CRCs of exported symbols are emitted as 32-bit relocations against
+	 * the NULL .dynsym entry, with the CRC value recorded in the addend.
+	 */
+6:	cmpdi	r0,R_PPC64_ADDR32
+	bne	7f
+	ld	r6,0(r9)	/* reloc->r_offset */
+	ld	r0,16(r9)	/* reloc->r_addend */
+	stwx	r0,r7,r6
+	addi	r9,r9,24
+
+7:	bdnz	5b
+	blr
 
-6:	blr
 
 .balign 8
 p_dyn:	.llong	__dynamic_start - 0b
diff --git a/arch/powerpc/relocs_check.sh b/arch/powerpc/relocs_check.sh
index ec2d5c835170..2f510fbc87da 100755
--- a/arch/powerpc/relocs_check.sh
+++ b/arch/powerpc/relocs_check.sh
@@ -43,7 +43,8 @@  R_PPC_ADDR16_HA
 R_PPC_RELATIVE
 R_PPC_NONE' |
 	grep -E -v '\<R_PPC64_ADDR64[[:space:]]+mach_' |
-	grep -E -v '\<R_PPC64_ADDR64[[:space:]]+__crc_'
+	grep -E -v '\<R_PPC64_ADDR64[[:space:]]+__crc_' |
+	grep -E -v '\<R_PPC64_ADDR32[[:space:]]+\*ABS\*'
 )
 
 if [ -z "$bad_relocs" ]; then