Re: [RFC PATCH v2 12/13] powerpc/kernel: Do not inconditionally save non volatile registers on system call
Christophe Leroy's on April 7, 2020 4:18 am: > > > Le 06/04/2020 à 03:25, Nicholas Piggin a écrit : >> Christophe Leroy's on April 6, 2020 3:44 am: >>> Before : 347 cycles on null_syscall >>> After : 327 cycles on null_syscall >> >> The problem I had doing this is that signal delivery wnats full regs, >> and you don't know if you have a signal pending ahead of time if you >> have interrupts enabled. >> >> I began to try bailing out back to asm to save nvgprs and call again. >> I think that can be made to work, but it is more complication in asm, >> and I soon found that 64s CPUs don't care about NVGPRs too much so it's >> nice to get rid of the !fullregs state. > > I tried a new way in v3, please have a look. I split > syscall_exit_prepare() in 3 parts and the result is unexpected: it is > better than before the series (307 cycles now versus 311 cycles with > full ASM syscall entry/exit). Great! Well I don't really see a problem with how you changed the C code around. I'll have to look at the assembly but I don't think it would have caused a problem for 64s. Thanks, Nick
Re: [RFC PATCH v2 12/13] powerpc/kernel: Do not inconditionally save non volatile registers on system call
Le 06/04/2020 à 03:25, Nicholas Piggin a écrit : Christophe Leroy's on April 6, 2020 3:44 am: Before : 347 cycles on null_syscall After : 327 cycles on null_syscall The problem I had doing this is that signal delivery wnats full regs, and you don't know if you have a signal pending ahead of time if you have interrupts enabled. I began to try bailing out back to asm to save nvgprs and call again. I think that can be made to work, but it is more complication in asm, and I soon found that 64s CPUs don't care about NVGPRs too much so it's nice to get rid of the !fullregs state. I tried a new way in v3, please have a look. I split syscall_exit_prepare() in 3 parts and the result is unexpected: it is better than before the series (307 cycles now versus 311 cycles with full ASM syscall entry/exit). Possibly another approach would be to leave interrupts disabled for the case where you have no work to do. You could create a small syscall_exit_prepare_nowork fastpath for that case for 32-bit, perhaps? Thanks, Nick Christophe
Re: [RFC PATCH v2 12/13] powerpc/kernel: Do not inconditionally save non volatile registers on system call
Christophe Leroy's on April 6, 2020 3:44 am: > Before : 347 cycles on null_syscall > After : 327 cycles on null_syscall The problem I had doing this is that signal delivery wnats full regs, and you don't know if you have a signal pending ahead of time if you have interrupts enabled. I began to try bailing out back to asm to save nvgprs and call again. I think that can be made to work, but it is more complication in asm, and I soon found that 64s CPUs don't care about NVGPRs too much so it's nice to get rid of the !fullregs state. Possibly another approach would be to leave interrupts disabled for the case where you have no work to do. You could create a small syscall_exit_prepare_nowork fastpath for that case for 32-bit, perhaps? Thanks, Nick
[RFC PATCH v2 12/13] powerpc/kernel: Do not inconditionally save non volatile registers on system call
Before : 347 cycles on null_syscall After : 327 cycles on null_syscall Signed-off-by: Christophe Leroy --- arch/powerpc/kernel/entry_32.S | 15 +++ arch/powerpc/kernel/head_32.h | 3 +-- arch/powerpc/kernel/syscall.c | 2 +- 3 files changed, 17 insertions(+), 3 deletions(-) diff --git a/arch/powerpc/kernel/entry_32.S b/arch/powerpc/kernel/entry_32.S index 103f5158bc44..b5113593e57f 100644 --- a/arch/powerpc/kernel/entry_32.S +++ b/arch/powerpc/kernel/entry_32.S @@ -315,13 +315,28 @@ stack_ovf: RFI #endif +save_nvgprs: + lwz r11, _TRAP(r1) + andi. r12, r11, 1 + rlwinm r11, r11, 0, ~1 + beqlr + SAVE_NVGPRS(r1) + stw r11, _TRAP(r1) + blr + .globl transfer_to_syscall transfer_to_syscall: + lwz r10, TI_FLAGS(r2) mr r9, r0 + andi. r10, r10, _TIF_SYSCALL_DOTRACE addir10, r1, STACK_FRAME_OVERHEAD + bnel- save_nvgprs bl system_call_exception ret_from_syscall: + lwz r9, TI_FLAGS(r2) addir4, r1, STACK_FRAME_OVERHEAD + andi. r0, r9, _TIF_SYSCALL_DOTRACE | _TIF_SINGLESTEP | _TIF_USER_WORK_MASK + bnel- save_nvgprs bl syscall_exit_prepare lwz r2, _CCR(r1) lwz r4, _NIP(r1) diff --git a/arch/powerpc/kernel/head_32.h b/arch/powerpc/kernel/head_32.h index c301d666a3e5..1cc9a67cb42c 100644 --- a/arch/powerpc/kernel/head_32.h +++ b/arch/powerpc/kernel/head_32.h @@ -174,13 +174,12 @@ END_MMU_FTR_SECTION_IFSET(MMU_FTR_HPTE_TABLE) stw r2,GPR2(r11) addir10,r10,STACK_FRAME_REGS_MARKER@l stw r9,_MSR(r11) - li r2, \trapno + li r2, \trapno + 1 stw r10,8(r11) stw r2,_TRAP(r11) SAVE_GPR(0, r11) SAVE_4GPRS(3, r11) SAVE_2GPRS(7, r11) - SAVE_NVGPRS(r11) addir11,r1,STACK_FRAME_OVERHEAD addir2,r12,-THREAD stw r11,PT_REGS(r12) diff --git a/arch/powerpc/kernel/syscall.c b/arch/powerpc/kernel/syscall.c index 630c423e089a..34fd66fd11a2 100644 --- a/arch/powerpc/kernel/syscall.c +++ b/arch/powerpc/kernel/syscall.c @@ -39,7 +39,7 @@ notrace long system_call_exception(long r3, long r4, long r5, if (!IS_ENABLED(CONFIG_PPC_BOOK3E)) BUG_ON(!(regs->msr & MSR_RI)); BUG_ON(IS_ENABLED(CONFIG_PPC64) && !(regs->msr & MSR_PR)); - BUG_ON(!FULL_REGS(regs)); + BUG_ON(IS_ENABLED(CONFIG_PPC64) && !FULL_REGS(regs)); #ifdef CONFIG_PPC64 BUG_ON(regs->softe != IRQS_ENABLED); #endif -- 2.25.0
