x86/vdso: Fix output operand size of RDPID insn
PATCH 1/2 x86/vdso: Fix output operand size of RDPID insn
RDPID instruction outputs to a word-sized register (64-bit on x86_64 and 32-bit on x86_32). Use unsigned long variable to use the correct size.
LSL outputs to 32-bit register, use %k operand prefix to always print 32-bit name of the register.
--- arch/x86/include/asm/segment.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/x86/include/asm/segment.h +2 -2
@@ -244,7 +244,7 @@ static inline unsigned long vdso_encode_cpunode(int cpu, unsigned long node)static inline void vdso_read_cpunode (unsigned *cpu,unsigned *node) { -unsigned int p; +unsigned long p;/* * Load CPU and node number from the GDT. LSL is faster than RDTSCP
@@ -254,7 +254,7 @@ static inline void vdso_read_cpunode(unsigned *cpu, unsigned *node)* * If RDPID is available, use it. */ -alternative_io ("lsl %[seg],%[p]" , +alternative_io ("lsl %[seg],%k[p]" ,".byte 0xf3,0x0f,0xc7,0xf8" ,/* RDPID %eax/rax */ X86_FEATURE_RDPID, [p] "=a" (p), [seg] "r" (__CPUNODE_SEG));
Signature · 1 line
2.49.0
PATCH 2/2 x86/vdso: Use RDPID mnemonic in vdso_red_cpunode()
Current minimum required version of binutils is 2.30, which supports RDPID instruction mnemonic.
Replace the byte-wise specification of RDPID with
this proper mnemonic.
No functional change intended.
--- arch/x86/include/asm/segment.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)
arch/x86/include/asm/segment.h +2 -2
@@ -255,9 +255,9 @@ static inline void vdso_read_cpunode(unsigned *cpu, unsigned *node)* If RDPID is available, use it. */ alternative_io ("lsl %[seg],%k[p]" , -".byte 0xf3,0x0f,0xc7,0xf8" ,/* RDPID %eax/rax */ +"rdpid %[p]" , X86_FEATURE_RDPID, - [p] "=a" (p), [seg] "r" (__CPUNODE_SEG)); + [p] "=r" (p), [seg] "r" (__CPUNODE_SEG)); if (cpu) *cpu = (p & VDSO_CPUNODE_MASK);
Signature · 1 line
2.49.0
[tip: x86/asm] x86/vdso: Fix output operand size of RDPID
In reply to Uros BizjakThe following commit has been merged into the x86/asm branch of tip:
The following commit has been merged into the x86/asm branch of tip:
Commit-ID: ac9c408ed19d535289ca59200dd6a44a6a2d6036 Gitweb: https://git.kernel.org/tip/ac9c408ed19d535289ca59200dd6a44a6a2d6036 Author: Uros Bizjak <[email protected]> AuthorDate: Mon, 16 Jun 2025 11:52:57 +02:00 Committer: Borislav Petkov (AMD) <[email protected]> CommitterDate: Tue, 26 Aug 2025 19:33:19 +02:00
x86/vdso: Fix output operand size of RDPID
RDPID instruction outputs to a word-sized register (64-bit on x86_64 and 32-bit on x86_32). Use an unsigned long variable to store the correct size.
LSL outputs to 32-bit register, use %k operand prefix to always print the 32-bit name of the register.
Use RDPID insn mnemonic while at it as the minimum binutils version of 2.30 supports it.
[ bp: Merge two patches touching the same function into a single one. ]
--- arch/x86/include/asm/segment.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)
arch/x86/include/asm/segment.h +4 -4
@@ -244,7 +244,7 @@ static inline unsigned long vdso_encode_cpunode(int cpu, unsigned long node)static inline void vdso_read_cpunode (unsigned *cpu,unsigned *node) { -unsigned int p; +unsigned long p;/* * Load CPU and node number from the GDT. LSL is faster than RDTSCP
@@ -254,10 +254,10 @@ static inline void vdso_read_cpunode(unsigned *cpu, unsigned *node)* * If RDPID is available, use it. */ -alternative_io ("lsl %[seg],%[p]" , -".byte 0xf3,0x0f,0xc7,0xf8" ,/* RDPID %eax/rax */ +alternative_io ("lsl %[seg],%k[p]" , +"rdpid %[p]" , X86_FEATURE_RDPID, - [p] "=a" (p), [seg] "r" (__CPUNODE_SEG)); + [p] "=r" (p), [seg] "r" (__CPUNODE_SEG)); if (cpu) *cpu = (p & VDSO_CPUNODE_MASK);
Re: [tip: x86/asm] x86/vdso: Fix output operand size of RDPID
Quoted 15 lines · On Wed, Aug 27, 2025 at 8:45 AM tip-bot2 for Uros Bizjak…
The following commit has been merged into the x86/asm branch of tip:
Commit-ID: ac9c408ed19d535289ca59200dd6a44a6a2d6036 Gitweb: https://git.kernel.org/tip/ac9c408ed19d535289ca59200dd6a44a6a2d6036 Author: Uros Bizjak <[email protected]> AuthorDate: Mon, 16 Jun 2025 11:52:57 +02:00 Committer: Borislav Petkov (AMD) <[email protected]> CommitterDate: Tue, 26 Aug 2025 19:33:19 +02:00x86/vdso: Fix output operand size of RDPID
RDPID instruction outputs to a word-sized register (64-bit on
x86_64and 32-bit onx86_32). Use an unsigned long variable to store the correct size.
Quoted 7 lines
LSL outputs to 32-bit register, use %k operand prefix to always print the 32-bit name of the register.
Use RDPID insn mnemonic while at it as the minimum binutils version of 2.30 supports it.
[ bp: Merge two patches touching the same function into a single one. ]
The 1/2 patch is intended to be backportable (only this part carries "Fixes" designation). The 2/2 patch introduces RDPID mnemonic that requires binutils-2.30, so the second part is not backportable to branches that still require binutils-2.25.
Uros.
Quoted 36 lines
Fixes ffebbaedc861 ("x86/vdso: Introduce helper functions for CPU and node number")Signed-off-by Uros Bizjak <[email protected]>Signed-off-by Borislav Petkov (AMD) <[email protected]>--- arch/x86/include/asm/segment.h | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-)arch/x86/include/asm/segment.h +4
-4@@ -244,7 +244,7 @@ static inline unsigned long vdso_encode_cpunode(int cpu, unsigned long node)static inline void vdso_read_cpunode (unsigned *cpu,unsigned *node) { -unsigned int p; +unsigned long p;/* * Load CPU and node number from the GDT. LSL is faster than RDTSCP @@ -254,10 +254,10 @@ static inline void vdso_read_cpunode(unsigned *cpu, unsigned *node)* * If RDPID is available, use it. */ -alternative_io ("lsl %[seg],%[p]" , -".byte 0xf3,0x0f,0xc7,0xf8" ,/* RDPID %eax/rax */ +alternative_io ("lsl %[seg],%k[p]" , +"rdpid %[p]" , X86_FEATURE_RDPID, - [p] "=a" (p), [seg] "r" (__CPUNODE_SEG)); + [p] "=r" (p), [seg] "r" (__CPUNODE_SEG)); if (cpu) *cpu = (p & VDSO_CPUNODE_MASK);
Borislav Petkov
maintainerRe: [tip: x86/asm] x86/vdso: Fix output operand size of RDPID
Quoted 1 line · Uros Bizjak
The 1/2 patch is intended to be backportable
Is it fixing an actual, real life bug?
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquetteRe: [tip: x86/asm] x86/vdso: Fix output operand size of RDPID
Quoted 5 lines · Borislav Petkov
Quoted 1 line · Uros Bizjak
The 1/2 patch is intended to be backportable
Is it fixing an actual, real life bug?
No, because ".byte 0xf3,0x0f,0xc7,0xf8" encodes "RDPID %rax" on 64-bit targets and we are only interested in the lower 32-bits anyway.
So, the change only matters on the 64-bit target when mnemonic is used (there is no RDPID %eax" instruction), but the patch should not be backported for the reasons outlined in my previous message. OTOH, "LSL %eax,%eax" zero-extends to the full 64-bit register width on 64-bit targets even when 32-bit register form is used.
Thanks,
Uros.