ARM: vfp: Use long jump to fix THUMB2 kernel compilation error
Subsystems (2)
PATCH 1/1 ARM: vfp: Use long jump to fix THUMB2 kernel compilation error
The vfp_kmode_exception() function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit: R_ARM_THM_JUMP19 against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.
--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
arch/arm/vfp/vfphw.S +2 -1
@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1
Signature · 1 line
2.27.0
Quoted 4 lines · Dmitry Osipenko
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.
Eek. Is this with gcc or clang?
Quoted 2 lines
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")
Are you sure it wasn't 512dd2eebe55 ("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0
arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1
This seems like a workaround though? I suspect the vfp11_veneer needs moving?
Signature · 1 line
Kees Cook
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.
I re-checked that the fixes tag is correct.
Quoted 17 lines
Quoted 17 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr
Quoted 6 lines
arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 2 lines
arch/arm/vfp/vfphw.S +0
-0VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
Quoted 43 lines · Dmitry Osipenko
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 17 lines
Quoted 17 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lrQuoted 6 lines
arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 2 lines
arch/arm/vfp/vfphw.S +0
-0VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
Signature · 1 line
Kees Cook
Quoted 45 lines · Kees Cook
Quoted 43 lines · Dmitry Osipenko
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?
Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
https://www.armlinux.org.uk/developer/patches/viewpatch.php?id=9018/1
Russell King - ARM Linux admin
maintainerQuoted 7 lines · Dmitry Osipenko
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Signed-off-by Dmitry Osipenko <[email protected]>
I guess the problem is that vfp_support_entry is in .text which tends to be at the beginning of the text section, but vfp_kmode_exception is in something like .text.vfp_kmode_exception ?
Would it be an idea just to change the section name that stuff like vfp_support_entry ends up in, rather than making the code less efficient?
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
22.10.2020 17:57, Russell King - ARM Linux admin пишет:
Quoted 6 lines · Dmitry Osipenko
Quoted 5 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.
Quoted 6 lines
Quoted 2 lines
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Signed-off-by Dmitry Osipenko <[email protected]>I guess the problem is that
vfp_support_entryis in .text which tends to be at the beginning of the text section, butvfp_kmode_exceptionis in something like .text.vfp_kmode_exception ?
Indeed, vfp_support_entry is at beginning of .text, while vfp_kmode_exception is near the end.
Quoted 4 lines
Would it be an idea just to change the section name that stuff like
vfp_support_entryends up in, rather than making the code less efficient?
This works:
arch/arm/vfp/vfphw.S +1 -0
@@ -65,6 +65,7 @@ #endif .endm + .pushsection .vfp11_veneer, "ax" @ VFP hardware support entry point. @
arch/arm/vfp/vfpmodule.c +1 -1
@@ -642,7 +642,7 @@ static int vfp_starting_cpu(unsigned int unused)return 0 ; } -void vfp_kmode_exception (void ) +void __section(".vfp11_veneer") vfp_kmode_exception(void) {/* * If we reach this point, a floating point exception has been raised
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 49 lines · Kees Cook
Quoted 48 lines · Dmitry Osipenko
Quoted 45 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 17 lines
Quoted 17 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lrQuoted 6 lines
arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 2 lines
arch/arm/vfp/vfphw.S +0
-0VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
Quoted 9 lines
Quoted 4 lines
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
https://www.armlinux.org.uk/developer/patches/viewpatch.php?id=9018/1
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Quoted 56 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 54 lines · Kees Cook
Quoted 52 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 17 lines
Quoted 17 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lrQuoted 6 lines
arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 2 lines
arch/arm/vfp/vfphw.S +0
-0VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
Quoted 7 lines
Quoted 4 lines
I already sent a fix for this issue:
https://www.armlinux.org.uk/developer/patches/viewpatch.php?id=9018/1
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Russell King - ARM Linux admin
maintainerQuoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 5 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
R_ARM_THM_JUMP19 only has -/+ 1 MB range, i.e., smaller than the size of .text, and so using it for calls to external symbols is still going to be risky, unless we create a completely separate code section for VFP related routines. So we may need the IT instruction anyway, at which point we don't need anything more.
Russell King - ARM Linux admin
maintainerQuoted 77 lines · Ard Biesheuvel
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Your patch may be a single line, but it has a slightly greater impact than the alternative two line solution.
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
Quoted 84 lines · Russell King - ARM Linux admin
Quoted 82 lines · Ard Biesheuvel
Quoted 79 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Quoted 8 lines
Quoted 3 lines
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Why should a short jump suffice? The call is to vfp_kmode_exception(), which we only call in exceptional cases. Why would we want to keep that in close proximity?
Quoted 3 lines
Your patch may be a single line, but it has a slightly greater impact than the alternative two line solution.
22.10.2020 19:23, Russell King - ARM Linux admin пишет:
Quoted 87 lines · Ard Biesheuvel
Quoted 77 lines
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
Quoted 7 lines
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Your patch may be a single line, but it has a slightly greater impact than the alternative two line solution.
But the two line change isn't portable to stable kernels as-is, isn't it?
Russell King - ARM Linux admin
maintainerQuoted 92 lines · Dmitry Osipenko
22.10.2020 19:23, Russell King - ARM Linux admin пишет:
Quoted 91 lines · Ard Biesheuvel
Quoted 77 lines
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Quoted 5 lines
Quoted 3 lines
Your patch may be a single line, but it has a slightly greater impact than the alternative two line solution.
But the two line change isn't portable to stable kernels as-is, isn't it?
Why not?
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
Quoted 94 lines · Russell King - ARM Linux admin
Quoted 92 lines · Dmitry Osipenko
22.10.2020 19:23, Russell King - ARM Linux admin пишет:
Quoted 91 lines · Ard Biesheuvel
Quoted 77 lines
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Quoted 8 lines
Quoted 5 lines
Quoted 3 lines
Your patch may be a single line, but it has a slightly greater impact than the alternative two line solution.
But the two line change isn't portable to stable kernels as-is, isn't it?
Why not?
In any case, I'd prefer not to dump VFP exception handling code into the .vfp11_veneer section, which is documented as below, and typically empty in our case, given that the only FP code we have in the kernel is NEON code.
"""
The '--vfp11-denorm-fix' switch enables a link-time workaround for a
bug in certain VFP11 coprocessor hardware, which sometimes allows instructions with denorm operands (which must be handled by support code) to have those operands overwritten by subsequent instructions before the support code can read the intended values.
The bug may be avoided in scalar mode if you allow at least one
intervening instruction between a VFP11 instruction which uses a register and another instruction which writes to the same register, or at least two intervening instructions if vector mode is in use. The bug only affects full-compliance floating-point mode: you do not need this workaround if you are using "runfast" mode. Please contact ARM for further details.
If you know you are using buggy VFP11 hardware, you can enable this
workaround by specifying the linker option '--vfp-denorm-fix=scalar' if you are using the VFP11 scalar mode only, or '--vfp-denorm-fix=vector' if you are using vector mode (the latter also works for scalar code). The default is '--vfp-denorm-fix=none'.
If the workaround is enabled, instructions are scanned for
potentially-troublesome sequences, and a veneer is created for each such sequence which may trigger the erratum. The veneer consists of the first instruction of the sequence and a branch back to the subsequent instruction. The original instruction is then replaced with a branch to the veneer. The extra cycles required to call and return from the veneer are sufficient to avoid the erratum in both the scalar and vector cases.
"""
Russell King - ARM Linux admin
maintainerQuoted 90 lines · Ard Biesheuvel
Quoted 88 lines · Russell King - ARM Linux admin
Quoted 85 lines · Ard Biesheuvel
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
Quoted 8 lines
Quoted 4 lines
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Why should a short jump suffice? The call is to
vfp_kmode_exception(), which we only call in exceptional cases. Why would we want to keep that in close proximity?
You're thinking about it in terms of what happens when the branch is taken, rather than also considering that this code path is also traversed for _every_ single time that we enter the support code not just for kernel mode.
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
Russell King - ARM Linux admin
maintainerQuoted 102 lines · Ard Biesheuvel
Quoted 100 lines · Russell King - ARM Linux admin
Quoted 97 lines · Dmitry Osipenko
22.10.2020 19:23, Russell King - ARM Linux admin пишет:
Quoted 87 lines · Ard Biesheuvel
Quoted 77 lines
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The vfp_kmode_exception() function now is unreachable using relative
branching in THUMB2 kernel configuration, resulting in a "relocation
truncated to fit: R_ARM_THM_JUMP19 against symbol `vfp_kmode_exception'"
linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes: eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")
Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 19 lines
Signed-off-by: Dmitry Osipenko <[email protected]>
---
arch/arm/vfp/vfphw.S | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)diff --git a/arch/arm/vfp/vfphw.S b/arch/arm/vfp/vfphw.S
index 4fcff9f59947..6e2b29f0c48d 100644
--- a/arch/arm/vfp/vfphw.S
+++ b/arch/arm/vfp/vfphw.S
@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry)
ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions
and r3, r3, #MODE_MASK @ are supported in kernel mode
teq r3, #USR_MODE
- bne vfp_kmode_exception @ Returns through lr
+ ldr r1, =vfp_kmode_exception
+ bxne r1 @ Returns through lrVFPFMRX r1, FPEXC @ Is the VFP enabled?
DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
Quoted 7 lines
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Your patch may be a single line, but it has a slightly greater impact than the alternative two line solution.
But the two line change isn't portable to stable kernels as-is, isn't it?
Quoted 7 lines
Quoted 2 lines
Why not?
In any case, I'd prefer not to dump VFP exception handling code into the .vfp11_veneer section, which is documented as below, and typically empty in our case, given that the only FP code we have in the kernel is NEON code.
This is getting out of hand, and really getting beyond a joke. I didn't say put it in the ".vfp11_veneer" section.
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/ FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
Quoted 107 lines · Russell King - ARM Linux admin
Quoted 105 lines · Ard Biesheuvel
Quoted 102 lines · Russell King - ARM Linux admin
Quoted 92 lines · Dmitry Osipenko
22.10.2020 19:23, Russell King - ARM Linux admin пишет:
Quoted 91 lines · Ard Biesheuvel
Quoted 77 lines
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
The vfp_kmode_exception() function now is unreachable using relative
branching in THUMB2 kernel configuration, resulting in a "relocation
truncated to fit: R_ARM_THM_JUMP19 against symbol `vfp_kmode_exception'"
linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Fixes: eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")
Are you sure it wasn't 512dd2eebe55 ("arm/build: Add missing sections") ?
That commit may have implicitly moved the location of .vfp11_veneer,
though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Signed-off-by: Dmitry Osipenko <[email protected]>
---
arch/arm/vfp/vfphw.S | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)diff --git a/arch/arm/vfp/vfphw.S b/arch/arm/vfp/vfphw.S
index 4fcff9f59947..6e2b29f0c48d 100644
--- a/arch/arm/vfp/vfphw.S
+++ b/arch/arm/vfp/vfphw.S
@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry)
ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions
and r3, r3, #MODE_MASK @ are supported in kernel mode
teq r3, #USR_MODE
- bne vfp_kmode_exception @ Returns through lr
+ ldr r1, =vfp_kmode_exception
+ bxne r1 @ Returns through lrVFPFMRX r1, FPEXC @ Is the VFP enabled?
DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the vfp11_veneer needs
moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Quoted 5 lines
Quoted 3 lines
Your patch may be a single line, but it has a slightly greater impact than the alternative two line solution.
But the two line change isn't portable to stable kernels as-is, isn't it?
Why not?
Quoted 8 lines
Quoted 4 lines
In any case, I'd prefer not to dump VFP exception handling code into the .vfp11_veneer section, which is documented as below, and typically empty in our case, given that the only FP code we have in the kernel is NEON code.
This is getting out of hand, and really getting beyond a joke. I didn't say put it in the ".vfp11_veneer" section.
No, but that is what Dmitry's patch proposes.
Quoted 97 lines · Russell King - ARM Linux admin
Quoted 95 lines · Ard Biesheuvel
Quoted 92 lines · Russell King - ARM Linux admin
Quoted 77 lines · Ard Biesheuvel
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The
vfp_kmode_exception()function now is unreachable using relative branching in THUMB2 kernel configuration, resulting in a "relocation truncated to fit:R_ARM_THM_JUMP19against symbol `vfp_kmode_exception'" linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 10 lines
Signed-off-by Dmitry Osipenko <[email protected]>--- arch/arm/vfp/vfphw.S | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)arch/arm/vfp/vfphw.S +0
-0arch/arm/vfp/vfphw.S ENTRY(vfp_support_entry) · line 89
Quoted 9 lines
arch/arm/vfp/vfphw.S +2
-1@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry) ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions and r3, r3, #MODE_MASK @ are supported in kernel mode teq r3, #USR_MODE - bne vfp_kmode_exception @ Returns through lr + ldr r1, =vfp_kmode_exception + bxne r1 @ Returns through lr VFPFMRX r1, FPEXC @ Is the VFP enabled? DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Quoted 9 lines
Quoted 3 lines
Why should a short jump suffice? The call is to
vfp_kmode_exception(), which we only call in exceptional cases. Why would we want to keep that in close proximity?You're thinking about it in terms of what happens when the branch is taken, rather than also considering that this code path is also traversed for _every_ single time that we enter the support code not just for kernel mode.
True. If 2 bytes of additional opcode are the concern here, we can change the current sequence
6: f093 0f10 teq r3, #16 a: f47f affe bne.w 0 <vfp_kmode_exception>
to
6: 2b10 cmp r3, #16 8: bf18 it ne a: f7ff bffe bne.w 0 <vfp_kmode_exception>
which takes up the exact same space.
22.10.2020 20:38, Russell King - ARM Linux admin пишет:
...
Quoted 4 lines
Quoted 1 line
But the two line change isn't portable to stable kernels as-is, isn't it?
Why not?
I assume we will need to define a new section, no?
Quoted 116 lines · Ard Biesheuvel
Quoted 97 lines · Russell King - ARM Linux admin
Quoted 95 lines · Ard Biesheuvel
Quoted 92 lines · Russell King - ARM Linux admin
Quoted 77 lines · Ard Biesheuvel
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
Quoted 4 lines
The vfp_kmode_exception() function now is unreachable using relative
branching in THUMB2 kernel configuration, resulting in a "relocation
truncated to fit: R_ARM_THM_JUMP19 against symbol `vfp_kmode_exception'"
linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Quoted 1 line
Fixes: eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")
Are you sure it wasn't
512dd2eebe55("arm/build: Add missing sections") ? That commit may have implicitly moved the location of .vfp11_veneer, though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Quoted 19 lines
Signed-off-by: Dmitry Osipenko <[email protected]>
---
arch/arm/vfp/vfphw.S | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)diff --git a/arch/arm/vfp/vfphw.S b/arch/arm/vfp/vfphw.S
index 4fcff9f59947..6e2b29f0c48d 100644
--- a/arch/arm/vfp/vfphw.S
+++ b/arch/arm/vfp/vfphw.S
@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry)
ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions
and r3, r3, #MODE_MASK @ are supported in kernel mode
teq r3, #USR_MODE
- bne vfp_kmode_exception @ Returns through lr
+ ldr r1, =vfp_kmode_exception
+ bxne r1 @ Returns through lrVFPFMRX r1, FPEXC @ Is the VFP enabled?
DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the
vfp11_veneerneeds moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Quoted 9 lines
Quoted 3 lines
Why should a short jump suffice? The call is to
vfp_kmode_exception(), which we only call in exceptional cases. Why would we want to keep that in close proximity?You're thinking about it in terms of what happens when the branch is taken, rather than also considering that this code path is also traversed for _every_ single time that we enter the support code not just for kernel mode.
True. If 2 bytes of additional opcode are the concern here, we can change the current sequence
6: f093 0f10 teq r3, #16 a: f47f affe bne.w 0 <vfp_kmode_exception>
Quoted 7 lines
to
6: 2b10 cmp r3, #16 8: bf18 it ne a: f7ff bffe bne.w 0 <vfp_kmode_exception>which takes up the exact same space.
BTW this code path looks slightly broken for Thumb-2 in any case: if a FP exception is taken in kernel mode on a Thumb2 kernel, we enter the emulation sequence via call_fpe, which will use the wrong set of value/mask pairs to match the opcode. The minimal fix is to move the call_fpe label to the right place, but I think it might be better to move the check for a FP exception in kernel mode to the handling of __und_svc.
Quoted 126 lines · Ard Biesheuvel
Quoted 116 lines · Ard Biesheuvel
Quoted 97 lines · Russell King - ARM Linux admin
Quoted 95 lines · Ard Biesheuvel
Quoted 92 lines · Russell King - ARM Linux admin
Quoted 77 lines · Ard Biesheuvel
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
Quoted 7 lines · Dmitry Osipenko
The vfp_kmode_exception() function now is unreachable using relative
branching in THUMB2 kernel configuration, resulting in a "relocation
truncated to fit: R_ARM_THM_JUMP19 against symbol `vfp_kmode_exception'"
linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Quoted 5 lines
Fixes: eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")
Are you sure it wasn't 512dd2eebe55 ("arm/build: Add missing sections") ?
That commit may have implicitly moved the location of .vfp11_veneer,
though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Quoted 23 lines
Signed-off-by: Dmitry Osipenko <[email protected]>
---
arch/arm/vfp/vfphw.S | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)diff --git a/arch/arm/vfp/vfphw.S b/arch/arm/vfp/vfphw.S
index 4fcff9f59947..6e2b29f0c48d 100644
--- a/arch/arm/vfp/vfphw.S
+++ b/arch/arm/vfp/vfphw.S
@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry)
ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions
and r3, r3, #MODE_MASK @ are supported in kernel mode
teq r3, #USR_MODE
- bne vfp_kmode_exception @ Returns through lr
+ ldr r1, =vfp_kmode_exception
+ bxne r1 @ Returns through lrVFPFMRX r1, FPEXC @ Is the VFP enabled?
DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the vfp11_veneer needs
moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a patch if you have a better idea, I'll be glad to test it.
I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Quoted 9 lines
Quoted 3 lines
Why should a short jump suffice? The call is to
vfp_kmode_exception(), which we only call in exceptional cases. Why would we want to keep that in close proximity?You're thinking about it in terms of what happens when the branch is taken, rather than also considering that this code path is also traversed for _every_ single time that we enter the support code not just for kernel mode.
True. If 2 bytes of additional opcode are the concern here, we can change the current sequence
6: f093 0f10 teq r3, #16 a: f47f affe bne.w 0 <vfp_kmode_exception>Quoted 7 lines
to
6: 2b10 cmp r3, #16 8: bf18 it ne a: f7ff bffe bne.w 0 <vfp_kmode_exception>which takes up the exact same space.
Quoted 7 lines
BTW this code path looks slightly broken for Thumb-2 in any case: if a FP exception is taken in kernel mode on a Thumb2 kernel, we enter the emulation sequence via
call_fpe, which will use the wrong set of value/mask pairs to match the opcode. The minimal fix is to move thecall_fpelabel to the right place, but I think it might be better to move the check for a FP exception in kernel mode to the handling of__und_svc.
Do we have a resolution here? This is causing breakage in kernelci
Quoted 136 lines · Ard Biesheuvel
Quoted 126 lines · Ard Biesheuvel
Quoted 116 lines · Ard Biesheuvel
Quoted 97 lines · Russell King - ARM Linux admin
Quoted 95 lines · Ard Biesheuvel
Quoted 92 lines · Russell King - ARM Linux admin
Quoted 77 lines · Ard Biesheuvel
Quoted 75 lines · Russell King - ARM Linux admin
Quoted 59 lines · Ard Biesheuvel
Quoted 58 lines · Dmitry Osipenko
22.10.2020 10:06, Ard Biesheuvel пишет:
Quoted 56 lines · Kees Cook
Quoted 45 lines · Dmitry Osipenko
Quoted 43 lines
22.10.2020 02:40, Kees Cook пишет:
On Thu, Oct 22, 2020 at 01:57:37AM +0300, Dmitry Osipenko wrote:
The vfp_kmode_exception() function now is unreachable using relative
branching in THUMB2 kernel configuration, resulting in a "relocation
truncated to fit: R_ARM_THM_JUMP19 against symbol `vfp_kmode_exception'"
linker error. Let's use long jump in order to fix the issue.Eek. Is this with gcc or clang?
GCC 9.3.0
Fixes: eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")
Are you sure it wasn't 512dd2eebe55 ("arm/build: Add missing sections") ?
That commit may have implicitly moved the location of .vfp11_veneer,
though I thought I had chosen the correct position.I re-checked that the fixes tag is correct.
Signed-off-by: Dmitry Osipenko <[email protected]>
---
arch/arm/vfp/vfphw.S | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)diff --git a/arch/arm/vfp/vfphw.S b/arch/arm/vfp/vfphw.S
index 4fcff9f59947..6e2b29f0c48d 100644
--- a/arch/arm/vfp/vfphw.S
+++ b/arch/arm/vfp/vfphw.S
@@ -82,7 +82,8 @@ ENTRY(vfp_support_entry)
ldr r3, [sp, #S_PSR] @ Neither lazy restore nor FP exceptions
and r3, r3, #MODE_MASK @ are supported in kernel mode
teq r3, #USR_MODE
- bne vfp_kmode_exception @ Returns through lr
+ ldr r1, =vfp_kmode_exception
+ bxne r1 @ Returns through lrVFPFMRX r1, FPEXC @ Is the VFP enabled?
DBGSTR1 "fpexc %08x", r1This seems like a workaround though? I suspect the vfp11_veneer needs
moving?Quoted 7 lines
Quoted 2 lines
I don't know where it needs to be moved. Please feel free to make a
patch if you have a better idea, I'll be glad to test it.I might have just been distracted by the common "vfp" prefix. It's possible that the text section shuffling just ended up being very large, so probably this patch is right then!
I already sent a fix for this issue:
Quoted 8 lines
Quoted 5 lines
The offending commit contains stable tag, so I assume that fixes tag is mandatory. Yours patch misses the fixes tag.
Russell, mind adding that? Or would you like me to update the patch in the patch system?
Rather than adding the IT, I'm suggesting that we solve it a different way - ensuring that the two bits of code are co-located. There's no reason for them to be separated, and the assembly code entry point is already called indirectly.
Quoted 8 lines
Quoted 4 lines
The problem is the assembly ends up in the .text section which ends up at the start of the binary, but depending on the compiler, functions in .c files end up in their own sections. It would be good if, as Dmitry has shown that it is indeed possible, to have them co-located.
Why is that better? I provided a minimal fix which has zero impact on ARM builds, and minimal impact on Thumb2 builds, given that it retains the exact same semantics as before, but using a different opcode.
I think you just described the reason there. Why should we force everything to use a different opcode when a short jump _should_ suffice?
Quoted 9 lines
Quoted 3 lines
Why should a short jump suffice? The call is to
vfp_kmode_exception(), which we only call in exceptional cases. Why would we want to keep that in close proximity?You're thinking about it in terms of what happens when the branch is taken, rather than also considering that this code path is also traversed for _every_ single time that we enter the support code not just for kernel mode.
True. If 2 bytes of additional opcode are the concern here, we can change the current sequence
6: f093 0f10 teq r3, #16 a: f47f affe bne.w 0 <vfp_kmode_exception>Quoted 7 lines
to
6: 2b10 cmp r3, #16 8: bf18 it ne a: f7ff bffe bne.w 0 <vfp_kmode_exception>which takes up the exact same space.
Quoted 7 lines
BTW this code path looks slightly broken for Thumb-2 in any case: if a FP exception is taken in kernel mode on a Thumb2 kernel, we enter the emulation sequence via
call_fpe, which will use the wrong set of value/mask pairs to match the opcode. The minimal fix is to move thecall_fpelabel to the right place, but I think it might be better to move the check for a FP exception in kernel mode to the handling of__und_svc.
Quoted 3 lines
Do we have a resolution here? This is causing breakage in kernelci
Still broken today
https://kernelci.org/build/id/5fa0c1a74bdb1ea4063fe7e4/
So the options are
a) merge my patch that adds 2 bytes of opcode to the Thumb2 build
b) merge Dmitry's patch that adds an unconditional literal load to all builds
c) remove kernel mode handling from vfp_support_entry() [my other patch]
d) move sections around so that vfp_kmode_exception is guaranteed to be in range.
e) do nothing
Given the lack of reports about this issue, it is pretty clear that few people use the Thumb2 build (which I find odd, tbh, since it really is much smaller). However, that means that a) is a reasonable fix, since nobody will notice the potential performance hit either, and it can easily be backported to wherever the breakage was introduced. (Note that eff8728fe698, which created the problem is marked cc:stable itself).
Going forward, I can refine d) so that we can get rid of the kernel mode path entirely.
03.11.2020 10:24, Ard Biesheuvel пишет:
Quoted 13 lines
Still broken today
https://kernelci.org/build/id/5fa0c1a74bdb1ea4063fe7e4/
So the options are
a) merge my patch that adds 2 bytes of opcode to the Thumb2 build
b) merge Dmitry's patch that adds an unconditional literal load to all builds
c) remove kernel mode handling fromvfp_support_entry()[my other patch]
d) move sections around so thatvfp_kmode_exceptionis guaranteed to be in range.
e) do nothing
Quoted 3 lines
Given the lack of reports about this issue, it is pretty clear that few people use the Thumb2 build (which I find odd, tbh, since it really is much smaller).
I waited for about a month, hoping that somebody will fix this problem before bothering with bisection, which took quite some time and effort because intermediate commits were broken, and then with creating and sending a patch :)
Thumb2 usually is untested by CI farms and in a case of personal use it's easier to wait for a fix. Hence no much reports, I suppose.
Quoted 5 lines
However, that means that a) is a reasonable
fix, since nobody will notice the potential performance hit either, and it can easily be backported to wherever the breakage was introduced. (Note thateff8728fe698, which created the problem is marked cc:stable itself).
The performance argument is questionable to me, to be honest. In practice the performance difference should be absolutely negligible for either of the proposed options, it should stay in a noise even if somebody thoroughly counting cycles, IMO.
I'm still thinking that the best option will be to apply a).
Quoted 2 lines
Going forward, I can refine d) so that we can get rid of the kernel mode path entirely.
And then improve it using d).
Quoted 6 lines · Dmitry Osipenko
03.11.2020 10:24, Ard Biesheuvel пишет:
Quoted 4 lines
Still broken today
Still broken today
https://kernelci.org/build/id/5fa898baa00b5f3167db8857/
Quoted 27 lines
Quoted 9 lines
So the options are
a) merge my patch that adds 2 bytes of opcode to the Thumb2 build
b) merge Dmitry's patch that adds an unconditional literal load to all builds
c) remove kernel mode handling fromvfp_support_entry()[my other patch]
d) move sections around so thatvfp_kmode_exceptionis guaranteed to be in range.
e) do nothingQuoted 3 lines
Given the lack of reports about this issue, it is pretty clear that few people use the Thumb2 build (which I find odd, tbh, since it really is much smaller).
I waited for about a month, hoping that somebody will fix this problem before bothering with bisection, which took quite some time and effort because intermediate commits were broken, and then with creating and sending a patch :)
Thumb2 usually is untested by CI farms and in a case of personal use it's easier to wait for a fix. Hence no much reports, I suppose.
Quoted 5 lines
However, that means that a) is a reasonable
fix, since nobody will notice the potential performance hit either, and it can easily be backported to wherever the breakage was introduced. (Note thateff8728fe698, which created the problem is marked cc:stable itself).
Quoted 7 lines
The performance argument is questionable to me, to be honest. In practice the performance difference should be absolutely negligible for either of the proposed options, it should stay in a noise even if somebody thoroughly counting cycles, IMO.
I'm still thinking that the best option will be to apply a).
Can we take that as an acked-by?
Quoted 4 lines
Quoted 2 lines
Going forward, I can refine d) so that we can get rid of the kernel mode path entirely.
And then improve it using d).
09.11.2020 10:25, Ard Biesheuvel пишет:
...
Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
So the options are
a) merge my patch that adds 2 bytes of opcode to the Thumb2 build
b) merge Dmitry's patch that adds an unconditional literal load to all builds
c) remove kernel mode handling fromvfp_support_entry()[my other patch]
d) move sections around so thatvfp_kmode_exceptionis guaranteed to be in range.
e) do nothing
...
Quoted 9 lines
Quoted 7 lines
The performance argument is questionable to me, to be honest. In practice the performance difference should be absolutely negligible for either of the proposed options, it should stay in a noise even if somebody thoroughly counting cycles, IMO.
I'm still thinking that the best option will be to apply a).
Can we take that as an acked-by?
Are you asking me for the ack? I think this is a more appropriate question to Russel. I'm not arm/ maintainer, but could give r-b and t-b.
If you're going to follow approach that I'm suggesting with a) + d), then could you please resend the two patches in a single series? The first one-line patch-fix should contain the fixes tag.
Quoted 17 lines · Dmitry Osipenko
09.11.2020 10:25, Ard Biesheuvel пишет:
...Quoted 8 lines
Quoted 8 lines
Quoted 8 lines
So the options are
a) merge my patch that adds 2 bytes of opcode to the Thumb2 build
b) merge Dmitry's patch that adds an unconditional literal load to all builds
c) remove kernel mode handling fromvfp_support_entry()[my other patch]
d) move sections around so thatvfp_kmode_exceptionis guaranteed to be in range.
e) do nothing...
Quoted 5 lines
Quoted 5 lines
The performance argument is questionable to me, to be honest. In practice the performance difference should be absolutely negligible for either of the proposed options, it should stay in a noise even if somebody thoroughly counting cycles, IMO.
Quoted 6 lines
Quoted 4 lines
Quoted 2 lines
I'm still thinking that the best option will be to apply a).
Can we take that as an acked-by?
Are you asking me for the ack?
Yes.
Quoted 3 lines
I think this is a more appropriate
question to Russel. I'm not arm/ maintainer, but could give r-b and t-b.
R-b and acked-by are basically the same thing.
Quoted 3 lines
If you're going to follow approach that I'm suggesting with a) + d), then could you please resend the two patches in a single series? The first one-line patch-fix should contain the fixes tag.
Ok