x86 -tip: entry_64.S cleanup series
Hi, here is a trivial cleanups for entry_64.S,
please review. The dangerous one is patch 2: get rid of back jump -- check it twice.
Alexander, I hope it doesn't interfere with the stuff you are working on?
- Cyrill -
PATCH 1/5 x86: entry_64.S - use ENTRY to define child_rip
From: Cyrill Gorcunov <[email protected]>
child_rip is called not by its name but indirectly rather so make it global and aligned.
--- arch/x86/kernel/entry_64.S | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
arch/x86/kernel/entry_64.S +1 -1
@@ -1322,7 +1322,7 @@ ENTRY(kernel_thread) CFI_ENDPROC END(kernel_thread) -child_rip: +ENTRY(child_rip) pushq $0 # fake return address CFI_STARTPROC /*
Signature · 1 line
1.6.0.4.603.gbc9c0
PATCH 2/5 x86: ret_from_fork - get rid of jump back
From: Cyrill Gorcunov <[email protected]>
Impact: cleanup, microoptimization
Use straightforward jumping if possible, cpu like it.
--- arch/x86/kernel/entry_64.S | 10 ++++------ 1 files changed, 4 insertions(+), 6 deletions(-)
arch/x86/kernel/entry_64.S +4 -6
@@ -379,7 +379,10 @@ ENTRY(ret_from_fork) GET_THREAD_INFO(%rcx) testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT),TI_flags(%rcx) CFI_REMEMBER_STATE - jnz rff_trace + jz rff_action + movq %rsp,%rdi + call syscall_trace_leave + GET_THREAD_INFO(%rcx) rff_action: RESTORE_REST testl $3,CS-ARGOFFSET(%rsp) # from kernel_thread?
@@ -389,11 +392,6 @@ rff_action: RESTORE_TOP_OF_STACK %rdi, -ARGOFFSET jmp ret_from_sys_call CFI_RESTORE_STATE -rff_trace: - movq %rsp,%rdi - call syscall_trace_leave - GET_THREAD_INFO(%rcx) - jmp rff_action CFI_ENDPROC END(ret_from_fork)
Signature · 1 line
1.6.0.4.603.gbc9c0
PATCH 3/5 x86: entry_64.S - use X86_EFLAGS_IF instead of hardcoded number
From: Cyrill Gorcunov <[email protected]>
Impact: cleanup
--- arch/x86/kernel/entry_64.S | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
arch/x86/kernel/entry_64.S +1 -1
@@ -199,7 +199,7 @@ ENTRY(native_usergs_sysret64) pushq %rax /* rsp */ CFI_ADJUST_CFA_OFFSET 8 CFI_REL_OFFSET rsp,0 - pushq $(1<<9) /* eflags - interrupts on */ + pushq $X86_EFLAGS_IF /* eflags - interrupts on */ CFI_ADJUST_CFA_OFFSET 8 /*CFI_REL_OFFSET rflags,0*/ pushq $__KERNEL_CS /* cs */
Signature · 1 line
1.6.0.4.603.gbc9c0
PATCH 4/5 x86: ret_from_fork - add CFI proc annotation
From: Cyrill Gorcunov <[email protected]>
ret_from_fork is definitly procedure so mark it as that.
--- arch/x86/kernel/entry_64.S | 1 + 1 files changed, 1 insertions(+), 0 deletions(-)
arch/x86/kernel/entry_64.S +1 -0
@@ -370,6 +370,7 @@ END(save_paranoid) */ /* rdi: prev */ ENTRY(ret_from_fork) + CFI_STARTPROC DEFAULT_FRAME push kernel_eflags(%rip) CFI_ADJUST_CFA_OFFSET 8
Signature · 1 line
1.6.0.4.603.gbc9c0
PATCH 5/5 x86: entry_64.S - trivial: space, comments fixup
From: Cyrill Gorcunov <[email protected]>
Impact: cleanup
--- arch/x86/kernel/entry_64.S | 92 ++++++++++++++++++++++--------------------- 1 files changed, 47 insertions(+), 45 deletions(-)
arch/x86/kernel/entry_64.S +47 -45
@@ -1119,8 +1119,8 @@ paranoidzeroentry machine_check do_machine_check zeroentry simd_coprocessor_error do_simd_coprocessor_error /* - * "Paranoid" exit path from exception stack. - * Paranoid because this is used by NMIs and cannot take + * "Paranoid" exit path from exception stack. + * Paranoid because this is used by NMIs and cannot take * any kernel state for granted. * We don't do kernel preemption checks here, because only * NMI should be common and it does not enable IRQs and
@@ -1225,7 +1225,7 @@ error_kernelspace: cmpq %rcx,RIP+8(%rsp) je error_swapgs cmpq $gs_change,RIP+8(%rsp) - je error_swapgs + je error_swapgs jmp error_sti KPROBE_END(error_entry)
@@ -1249,36 +1249,36 @@ KPROBE_ENTRY(error_exit) CFI_ENDPROC KPROBE_END(error_exit) - /* Reload gs selector with exception handling */ - /* edi: new selector */ + /* Reload gs selector with exception handling */ + /* edi: new selector */ ENTRY(native_load_gs_index) CFI_STARTPROC pushf CFI_ADJUST_CFA_OFFSET 8 DISABLE_INTERRUPTS(CLBR_ANY | ~(CLBR_RDI)) - SWAPGS + SWAPGS gs_change: - movl %edi,%gs + movl %edi,%gs 2: mfence /* workaround */ SWAPGS - popf + popf CFI_ADJUST_CFA_OFFSET -8 - ret + ret CFI_ENDPROC END(native_load_gs_index) - .section __ex_table,"a" - .align 8 - .quad gs_change,bad_gs - .previous - .section .fixup,"ax" + .section __ex_table,"a" + .align 8 + .quad gs_change,bad_gs + .previous + .section .fixup,"ax" /* running with kernelgs */ bad_gs: SWAPGS /* switch back to user gs */ xorl %eax,%eax - movl %eax,%gs - jmp 2b - .previous + movl %eax,%gs + jmp 2b + .previous /* * Create a kernel thread.
@@ -1313,7 +1313,7 @@ ENTRY(kernel_thread) * so internally to the x86_64 port you can rely on kernel_thread() * not to reschedule the child before returning, this avoids the need * of hacks for example to fork off the per-CPU idle tasks. - * [Hopefully no generic code relies on the reschedule -AK] + * [Hopefully no generic code relies on the reschedule -AK] */ RESTORE_ALL UNFAKE_STACK_FRAME
@@ -1420,7 +1420,7 @@ nmi_schedule: CFI_ENDPROC #else jmp paranoid_exit - CFI_ENDPROC + CFI_ENDPROC #endif KPROBE_END(nmi)
@@ -1455,22 +1455,24 @@ KPROBE_END(ignore_sysret) zeroentry xen_hypervisor_callback xen_do_hypervisor_callback /* -# A note on the "critical region" in our callback handler. -# We want to avoid stacking callback handlers due to events occurring -# during handling of the last event. To do this, we keep events disabled -# until we've done all processing. HOWEVER, we must enable events before -# popping the stack frame (can't be done atomically) and so it would still -# be possible to get enough handler activations to overflow the stack. -# Although unlikely, bugs of that kind are hard to track down, so we'd -# like to avoid the possibility. -# So, on entry to the handler we detect whether we interrupted an -# existing activation in its critical region -- if so, we pop the current -# activation and restart the handler using the previous one. -*/ + * A note on the "critical region" in our callback handler. + * We want to avoid stacking callback handlers due to events occurring + * during handling of the last event. To do this, we keep events disabled + * until we've done all processing. HOWEVER, we must enable events before + * popping the stack frame (can't be done atomically) and so it would still + * be possible to get enough handler activations to overflow the stack. + * Although unlikely, bugs of that kind are hard to track down, so we'd + * like to avoid the possibility. + * So, on entry to the handler we detect whether we interrupted an + * existing activation in its critical region -- if so, we pop the current + * activation and restart the handler using the previous one. + */ ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs) CFI_STARTPROC -/* Since we don't modify %rdi, evtchn_do_upall(struct *pt_regs) will - see the correct pointer to the pt_regs */ +/* + * Since we don't modify %rdi, evtchn_do_upall(struct *pt_regs) will + * see the correct pointer to the pt_regs + */ movq %rdi, %rsp # we don't return, adjust the stack frame CFI_ENDPROC DEFAULT_FRAME
@@ -1488,18 +1490,18 @@ ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs) END(do_hypervisor_callback) /* -# Hypervisor uses this for application faults while it executes. -# We get here for two reasons: -# 1. Fault while reloading DS, ES, FS or GS -# 2. Fault while executing IRET -# Category 1 we do not need to fix up as Xen has already reloaded all segment -# registers that could be reloaded and zeroed the others. -# Category 2 we fix up by killing the current process. We cannot use the -# normal Linux return path in this case because if we use the IRET hypercall -# to pop the stack frame we end up in an infinite loop of failsafe callbacks. -# We distinguish between categories by comparing each saved segment register -# with its current contents: any discrepancy means we in category 1. -*/ + * Hypervisor uses this for application faults while it executes. + * We get here for two reasons: + * 1. Fault while reloading DS, ES, FS or GS + * 2. Fault while executing IRET + * Category 1 we do not need to fix up as Xen has already reloaded all segment + * registers that could be reloaded and zeroed the others. + * Category 2 we fix up by killing the current process. We cannot use the + * normal Linux return path in this case because if we use the IRET hypercall + * to pop the stack frame we end up in an infinite loop of failsafe callbacks. + * We distinguish between categories by comparing each saved segment register + * with its current contents: any discrepancy means we in category 1. + */ ENTRY(xen_failsafe_callback) INTR_FRAME 1 (6*8) /*CFI_REL_OFFSET gs,GS*/
Signature · 1 line
1.6.0.4.603.gbc9c0
Re: [PATCH 5/5] x86: entry_64.S - trivial: space, comments fixup
Quoted 5 lines · On Wed, 26 Nov 2008 22:17:04 +0300, [email protected] said:
From: Cyrill Gorcunov <[email protected]>
Impact: cleanup
Signed-off-by Cyrill Gorcunov <[email protected]>
Quoted 187 lines
arch/x86/kernel/entry_64.S | 92 ++++++++++++++++++++++--------------------- 1 files changed, 47 insertions(+), 45 deletions(-)arch/x86/kernel/entry_64.S +19
-24@@ -1119,8 +1119,8 @@ paranoidzeroentry machine_check do_machine_check zeroentry simd_coprocessor_error do_simd_coprocessor_error /* - * "Paranoid" exit path from exception stack. - * Paranoid because this is used by NMIs and cannot take + * "Paranoid" exit path from exception stack. + * Paranoid because this is used by NMIs and cannot take * any kernel state for granted. * We don't do kernel preemption checks here, because only * NMI should be common and it does not enable IRQs and@@ -1225,7 +1225,7 @@ error_kernelspace: cmpq %rcx,RIP+8(%rsp) je error_swapgs cmpq $gs_change,RIP+8(%rsp) - je error_swapgs + je error_swapgs jmp error_sti KPROBE_END(error_entry)@@ -1249,36 +1249,36 @@ KPROBE_ENTRY(error_exit) CFI_ENDPROC KPROBE_END(error_exit) - /* Reload gs selector with exception handling */ - /* edi: new selector */ + /* Reload gs selector with exception handling */ + /* edi: new selector */ ENTRY(native_load_gs_index) CFI_STARTPROC pushf CFI_ADJUST_CFA_OFFSET 8 DISABLE_INTERRUPTS(CLBR_ANY | ~(CLBR_RDI)) - SWAPGS + SWAPGS gs_change: - movl %edi,%gs + movl %edi,%gs 2: mfence /* workaround */ SWAPGS - popf + popf CFI_ADJUST_CFA_OFFSET -8 - ret + ret CFI_ENDPROC END(native_load_gs_index) - .section __ex_table,"a" - .align 8 - .quad gs_change,bad_gs - .previous - .section .fixup,"ax" + .section __ex_table,"a" + .align 8 + .quad gs_change,bad_gs + .previous + .section .fixup,"ax" /* running with kernelgs */ bad_gs: SWAPGS /* switch back to user gs */ xorl %eax,%eax - movl %eax,%gs - jmp 2b - .previous + movl %eax,%gs + jmp 2b + .previous /* * Create a kernel thread.@@ -1313,7 +1313,7 @@ ENTRY(kernel_thread) * so internally to the x86_64 port you can rely on kernel_thread() * not to reschedule the child before returning, this avoids the need * of hacks for example to fork off the per-CPU idle tasks. - * [Hopefully no generic code relies on the reschedule -AK] + * [Hopefully no generic code relies on the reschedule -AK] */ RESTORE_ALL UNFAKE_STACK_FRAME@@ -1420,7 +1420,7 @@ nmi_schedule: CFI_ENDPROC #else jmp paranoid_exit - CFI_ENDPROC + CFI_ENDPROC #endif KPROBE_END(nmi)@@ -1455,22 +1455,24 @@ KPROBE_END(ignore_sysret) zeroentry xen_hypervisor_callback xen_do_hypervisor_callback /* -# A note on the "critical region" in our callback handler. -# We want to avoid stacking callback handlers due to events occurring -# during handling of the last event. To do this, we keep events disabled -# until we've done all processing. HOWEVER, we must enable events before -# popping the stack frame (can't be done atomically) and so it wouldstill
arch/x86/kernel/entry_64.S +0
-5-# be possible to get enough handler activations to overflow the stack. -# Although unlikely, bugs of that kind are hard to track down, so we'd -# like to avoid the possibility. -# So, on entry to the handler we detect whether we interrupted an -# existing activation in its critical region -- if so, we pop thecurrent
arch/x86/kernel/entry_64.S +3
-2-# activation and restart the handler using the previous one. -*/ + * A note on the "critical region" in our callback handler. + * We want to avoid stacking callback handlers due to events occurring + * during handling of the last event. To do this, we keep eventsdisabled
+ * until we've done all processing. HOWEVER, we must enable events before
+ * popping the stack frame (can't be done atomically) and so it would still
+ * be possible to get enough handler activations to overflow the stack.
+ * Although unlikely, bugs of that kind are hard to track down, so we'd
+ * like to avoid the possibility.
+ * So, on entry to the handler we detect whether we interrupted an
+ * existing activation in its critical region -- if so, we pop the current
+ * activation and restart the handler using the previous one.
+ */
ENTRY(xen_do_hypervisor_callback)# do_hypervisor_callback(struct
*pt_regs)arch/x86/kernel/entry_64.S +4
-7CFI_STARTPROC -/* Since we don't modify %rdi, evtchn_do_upall(struct *pt_regs) will - see the correct pointer to the pt_regs */ +/* + * Since we don't modify %rdi, evtchn_do_upall(struct *pt_regs) will + * see the correct pointer to the pt_regs + */ movq %rdi, %rsp # we don't return, adjust the stack frame CFI_ENDPROC DEFAULT_FRAME@@ -1488,18 +1490,18 @@ ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs) END(do_hypervisor_callback) /* -# Hypervisor uses this for application faults while it executes. -# We get here for two reasons: -# 1. Fault while reloading DS, ES, FS or GS -# 2. Fault while executing IRET -# Category 1 we do not need to fix up as Xen has already reloaded allsegment
arch/x86/kernel/entry_64.S +0
-3-# registers that could be reloaded and zeroed the others. -# Category 2 we fix up by killing the current process. We cannot use the -# normal Linux return path in this case because if we use the IREThypercall
-# to pop the stack frame we end up in an infinite loop of failsafe callbacks.
-# We distinguish between categories by comparing each saved segment registerarch/x86/kernel/entry_64.S +5
-2-# with its current contents: any discrepancy means we in category 1. -*/ + * Hypervisor uses this for application faults while it executes. + * We get here for two reasons: + * 1. Fault while reloading DS, ES, FS or GS + * 2. Fault while executing IRET + * Category 1 we do not need to fix up as Xen has already reloaded allsegment
+ * registers that could be reloaded and zeroed the others.
+ * Category 2 we fix up by killing the current process. We cannot use the
+ * normal Linux return path in this case because if we use the IRET hypercall
+ * to pop the stack frame we end up in an infinite loop of failsafe callbacks.
+ * We distinguish between categories by comparing each saved segment register
+ * with its current contents: any discrepancy means we in category 1.
+ */
ENTRY(xen_failsafe_callback)arch/x86/kernel/entry_64.S +0
-0INTR_FRAME 1 (6*8) /*CFI_REL_OFFSET gs,GS*/Signature · 1 line
1.6.0.4.603.gbc9c0
Signature · 5 lines
Alexander van Heukelum [email protected] -- http://www.fastmail.fm - The way an email service should be
Re: [PATCH 5/5] x86: entry_64.S - trivial: space, comments fixup
Quoted 8 lines · Alexander van Heukelum
Quoted 5 lines · On Wed, 26 Nov 2008 22:17:04 +0300, [email protected] said:
From: Cyrill Gorcunov <[email protected]>
Impact: cleanup
Signed-off-by Cyrill Gorcunov <[email protected]>
Cyrill, could you please rework the still relevant bits against latest tip/master? Alexander's wider patch got in the way.
Thanks,
Ingo
Re: [PATCH 5/5] x86: entry_64.S - trivial: space, comments fixup
Quoted 10 lines · Ingo Molnar
Quoted 7 lines · Alexander van Heukelum
Quoted 5 lines · On Wed, 26 Nov 2008 22:17:04 +0300, [email protected] said:
From: Cyrill Gorcunov <[email protected]>
Impact: cleanup
Signed-off-by Cyrill Gorcunov <[email protected]>
Quoted 9 lines
Quoted 1 line
Cyrill, could you please rework the still relevant bits against latest tip/master? Alexander's wider patch got in the way.
Thanks,
Ingo
yep, no problem (wait 'till evening please)
Re: [PATCH 5/5] x86: entry_64.S - trivial: space, comments fixup
In reply to Ingo Molnar[Ingo Molnar - Thu, Nov 27, 2008 at 01:02:34PM +0100]
| > > From: Cyrill Gorcunov <[email protected]> | > > | > > Impact: cleanup | > > | > > Signed-off-by: Cyrill Gorcunov <[email protected]> | > | > Acked-by: Alexander van Heukelum <[email protected]> | | Cyrill, could you please rework the still relevant bits against latest | tip/master? Alexander's wider patch got in the way. | | Thanks, | | Ingo |
Here is an updated version
---
From: Cyrill Gorcunov <[email protected]>
Subject: x86: entry_64.S - trivial: space, comments fixup
Impact: cleanup
--- arch/x86/kernel/entry_64.S | 94 ++++++++++++++++++++++----------------------- 1 file changed, 48 insertions(+), 46 deletions(-)
arch/x86/kernel/entry_64.S +48 -46
@@ -1027,7 +1027,7 @@ END(\sym) .macro paranoidzeroentry_ist sym do_sym ist ENTRY(\sym) - INTR_FRAME + INTR_FRAME PARAVIRT_ADJUST_EXCEPTION_FRAME pushq $-1 /* ORIG_RAX: no syscall to restart */ CFI_ADJUST_CFA_OFFSET 8
@@ -1095,36 +1095,36 @@ zeroentry coprocessor_error do_coprocess errorentry alignment_check do_alignment_check zeroentry simd_coprocessor_error do_simd_coprocessor_error - /* Reload gs selector with exception handling */ - /* edi: new selector */ + /* Reload gs selector with exception handling */ + /* edi: new selector */ ENTRY(native_load_gs_index) CFI_STARTPROC pushf CFI_ADJUST_CFA_OFFSET 8 DISABLE_INTERRUPTS(CLBR_ANY | ~(CLBR_RDI)) - SWAPGS + SWAPGS gs_change: - movl %edi,%gs + movl %edi,%gs 2: mfence /* workaround */ SWAPGS - popf + popf CFI_ADJUST_CFA_OFFSET -8 - ret + ret CFI_ENDPROC END(native_load_gs_index) - .section __ex_table,"a" - .align 8 - .quad gs_change,bad_gs - .previous - .section .fixup,"ax" + .section __ex_table,"a" + .align 8 + .quad gs_change,bad_gs + .previous + .section .fixup,"ax" /* running with kernelgs */ bad_gs: SWAPGS /* switch back to user gs */ xorl %eax,%eax - movl %eax,%gs - jmp 2b - .previous + movl %eax,%gs + jmp 2b + .previous /* * Create a kernel thread.
@@ -1159,7 +1159,7 @@ ENTRY(kernel_thread) * so internally to the x86_64 port you can rely on kernel_thread() * not to reschedule the child before returning, this avoids the need * of hacks for example to fork off the per-CPU idle tasks. - * [Hopefully no generic code relies on the reschedule -AK] + * [Hopefully no generic code relies on the reschedule -AK] */ RESTORE_ALL UNFAKE_STACK_FRAME
@@ -1239,22 +1239,24 @@ END(call_softirq) zeroentry xen_hypervisor_callback xen_do_hypervisor_callback /* -# A note on the "critical region" in our callback handler. -# We want to avoid stacking callback handlers due to events occurring -# during handling of the last event. To do this, we keep events disabled -# until we've done all processing. HOWEVER, we must enable events before -# popping the stack frame (can't be done atomically) and so it would still -# be possible to get enough handler activations to overflow the stack. -# Although unlikely, bugs of that kind are hard to track down, so we'd -# like to avoid the possibility. -# So, on entry to the handler we detect whether we interrupted an -# existing activation in its critical region -- if so, we pop the current -# activation and restart the handler using the previous one. -*/ + * A note on the "critical region" in our callback handler. + * We want to avoid stacking callback handlers due to events occurring + * during handling of the last event. To do this, we keep events disabled + * until we've done all processing. HOWEVER, we must enable events before + * popping the stack frame (can't be done atomically) and so it would still + * be possible to get enough handler activations to overflow the stack. + * Although unlikely, bugs of that kind are hard to track down, so we'd + * like to avoid the possibility. + * So, on entry to the handler we detect whether we interrupted an + * existing activation in its critical region -- if so, we pop the current + * activation and restart the handler using the previous one. + */ ENTRY(xen_do_hypervisor_callback) # do_hypervisor_callback(struct *pt_regs) CFI_STARTPROC -/* Since we don't modify %rdi, evtchn_do_upall(struct *pt_regs) will - see the correct pointer to the pt_regs */ +/* + * Since we don't modify %rdi, evtchn_do_upall(struct *pt_regs) will + * see the correct pointer to the pt_regs + */ movq %rdi, %rsp # we don't return, adjust the stack frame CFI_ENDPROC DEFAULT_FRAME
@@ -1272,18 +1274,18 @@ ENTRY(xen_do_hypervisor_callback) # do END(do_hypervisor_callback) /* -# Hypervisor uses this for application faults while it executes. -# We get here for two reasons: -# 1. Fault while reloading DS, ES, FS or GS -# 2. Fault while executing IRET -# Category 1 we do not need to fix up as Xen has already reloaded all segment -# registers that could be reloaded and zeroed the others. -# Category 2 we fix up by killing the current process. We cannot use the -# normal Linux return path in this case because if we use the IRET hypercall -# to pop the stack frame we end up in an infinite loop of failsafe callbacks. -# We distinguish between categories by comparing each saved segment register -# with its current contents: any discrepancy means we in category 1. -*/ + * Hypervisor uses this for application faults while it executes. + * We get here for two reasons: + * 1. Fault while reloading DS, ES, FS or GS + * 2. Fault while executing IRET + * Category 1 we do not need to fix up as Xen has already reloaded all segment + * registers that could be reloaded and zeroed the others. + * Category 2 we fix up by killing the current process. We cannot use the + * normal Linux return path in this case because if we use the IRET hypercall + * to pop the stack frame we end up in an infinite loop of failsafe callbacks. + * We distinguish between categories by comparing each saved segment register + * with its current contents: any discrepancy means we in category 1. + */ ENTRY(xen_failsafe_callback) INTR_FRAME 1 (6*8) /*CFI_REL_OFFSET gs,GS*/
@@ -1347,8 +1349,8 @@ paranoidzeroentry machine_check do_machi #endif /* - * "Paranoid" exit path from exception stack. - * Paranoid because this is used by NMIs and cannot take + * "Paranoid" exit path from exception stack. + * Paranoid because this is used by NMIs and cannot take * any kernel state for granted. * We don't do kernel preemption checks here, because only * NMI should be common and it does not enable IRQs and
@@ -1453,7 +1455,7 @@ error_kernelspace: cmpq %rcx,RIP+8(%rsp) je error_swapgs cmpq $gs_change,RIP+8(%rsp) - je error_swapgs + je error_swapgs jmp error_sti END(error_entry)
@@ -1529,7 +1531,7 @@ nmi_schedule: CFI_ENDPROC #else jmp paranoid_exit - CFI_ENDPROC + CFI_ENDPROC #endif END(nmi)
Ingo Molnar
AppliedRe: [PATCH 5/5] x86: entry_64.S - trivial: space, comments fixup
Quoted 17 lines · Cyrill Gorcunov
[Ingo Molnar - Thu, Nov 27, 2008 at 01:02:34PM +0100]
| > > From: Cyrill Gorcunov <[email protected]> | > > | > > Impact: cleanup | > > | > > Signed-off-by: Cyrill Gorcunov <[email protected]> | > | > Acked-by: Alexander van Heukelum <[email protected]> | | Cyrill, could you please rework the still relevant bits against latest | tip/master? Alexander's wider patch got in the way. | | Thanks, | | Ingo |
Quoted 6 lines
Here is an updated version
---
From: Cyrill Gorcunov <[email protected]>
Subject: x86:entry_64.S- trivial: space, comments fixupImpact: cleanup
applied, thanks Cyrill!
Ingo
Re: [PATCH 4/5] x86: ret_from_fork - add CFI proc annotation
In reply to [email protected]Quoted 14 lines · On Wed, 26 Nov 2008 22:17:03 +0300, [email protected] said:
From: Cyrill Gorcunov <[email protected]>
ret_from_forkis definitly procedure so mark it as that.Signed-off-by Cyrill Gorcunov <[email protected]>--- arch/x86/kernel/entry_64.S | 1 + 1 files changed, 1 insertions(+), 0 deletions(-)arch/x86/kernel/entry_64.S +0
-0
arch/x86/kernel/entry_64.S END(save_paranoid) · line 376
Quoted 7 lines
arch/x86/kernel/entry_64.S +1
-0@@ -370,6 +370,7 @@ END(save_paranoid) */ /* rdi: prev */ ENTRY(ret_from_fork) + CFI_STARTPROC DEFAULT_FRAME push kernel_eflags(%rip) CFI_ADJUST_CFA_OFFSET 8
This is not needed/wanted. DEFAULT_FRAME includes "CFI_STARTPROC simple" already.
Alexander
Quoted 2 lines
1.6.0.4.603.gbc9c0
Signature · 5 lines
Alexander van Heukelum [email protected] -- http://www.fastmail.fm - The professional email service
Re: [PATCH 4/5] x86: ret_from_fork - add CFI proc annotation
Quoted 29 lines · Alexander van Heukelum
Quoted 14 lines · On Wed, 26 Nov 2008 22:17:03 +0300, [email protected] said:
From: Cyrill Gorcunov <[email protected]>
ret_from_forkis definitly procedure so mark it as that.Signed-off-by Cyrill Gorcunov <[email protected]>--- arch/x86/kernel/entry_64.S | 1 + 1 files changed, 1 insertions(+), 0 deletions(-)arch/x86/kernel/entry_64.S +0
-0arch/x86/kernel/entry_64.S END(save_paranoid) · line 376
Quoted 7 lines
arch/x86/kernel/entry_64.S +1
-0@@ -370,6 +370,7 @@ END(save_paranoid) */ /* rdi: prev */ ENTRY(ret_from_fork) + CFI_STARTPROC DEFAULT_FRAME push kernel_eflags(%rip) CFI_ADJUST_CFA_OFFSET 8This is not needed/wanted.
DEFAULT_FRAMEincludes "CFI_STARTPROCsimple" already.Alexander
Quoted 10 lines
Quoted 2 lines
1.6.0.4.603.gbc9c0
--
Alexander van Heukelum
[email protected]--
http://www.fastmail.fm - The professional email service
Indeed, thanks (such an indirect and nested definition sometime confusing)
Re: [PATCH 3/5] x86: entry_64.S - use X86_EFLAGS_IF instead of hardcoded number
In reply to [email protected]Quoted 5 lines · On Wed, 26 Nov 2008 22:17:02 +0300, [email protected] said:
From: Cyrill Gorcunov <[email protected]>
Impact: cleanup
Signed-off-by Cyrill Gorcunov <[email protected]>
Quoted 19 lines
arch/x86/kernel/entry_64.S | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)arch/x86/kernel/entry_64.S +1
-1@@ -199,7 +199,7 @@ ENTRY(native_usergs_sysret64) pushq %rax /* rsp */ CFI_ADJUST_CFA_OFFSET 8 CFI_REL_OFFSET rsp,0 - pushq $(1<<9) /* eflags - interrupts on */ + pushq $X86_EFLAGS_IF /* eflags - interrupts on */ CFI_ADJUST_CFA_OFFSET 8 /*CFI_REL_OFFSET rflags,0*/ pushq $__KERNEL_CS /* cs */1.6.0.4.603.gbc9c0
Signature · 5 lines
Alexander van Heukelum [email protected] -- http://www.fastmail.fm - A no graphics, no pop-ups email service
Ingo Molnar
AppliedRe: [PATCH 3/5] x86: entry_64.S - use X86_EFLAGS_IF instead of hardcoded number
Quoted 8 lines · Alexander van Heukelum
Quoted 5 lines · On Wed, 26 Nov 2008 22:17:02 +0300, [email protected] said:
From: Cyrill Gorcunov <[email protected]>
Impact: cleanup
Signed-off-by Cyrill Gorcunov <[email protected]>
applied, thanks guys!
Ingo
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
In reply to [email protected][[email protected] - Wed, Nov 26, 2008 at 10:17:01PM +0300]
| From: Cyrill Gorcunov <[email protected]> | | Impact: cleanup, microoptimization | | Use straightforward jumping if possible, cpu like it.
Wrong word, not _straight_ forward but forward branch target in this case.
| | Signed-off-by: Cyrill Gorcunov <[email protected]> | --- | arch/x86/kernel/entry_64.S | 10 ++++------ | 1 files changed, 4 insertions(+), 6 deletions(-) |
- Cyrill -
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
In reply to [email protected]arch/x86/kernel/entry_64.S ENTRY(ret_from_fork) · line 385
Quoted 11 lines · [email protected] writes:
arch/x86/kernel/entry_64.S +4
-1@@ -379,7 +379,10 @@ ENTRY(ret_from_fork) GET_THREAD_INFO(%rcx) testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT),TI_flags(%rcx) CFI_REMEMBER_STATE - jnz rff_trace + jz rff_action + movq %rsp,%rdi + call syscall_trace_leave + GET_THREAD_INFO(%rcx)
The uncommon path is supposed to be out of line. I don't think the CPU will like that.
-Andi
Quoted 1 line
arch/x86/kernel/entry_64.S +0
-0rff_action:
Signature · 1 line
[email protected]
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
[Andi Kleen - Wed, Nov 26, 2008 at 09:04:33PM +0100]
| [email protected] writes: | > --- a/arch/x86/kernel/entry_64.S | > +++ b/arch/x86/kernel/entry_64.S | > @@ -379,7 +379,10 @@ ENTRY(ret_from_fork) | > GET_THREAD_INFO(%rcx) | > testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT),TI_flags(%rcx) | > CFI_REMEMBER_STATE | > - jnz rff_trace | > + jz rff_action | > + movq %rsp,%rdi | > + call syscall_trace_leave | > + GET_THREAD_INFO(%rcx) | | The uncommon path is supposed to be out of line. I don't think | the CPU will like that. | | -Andi | | > rff_action: | | -- | [email protected] |
Aha! Thanks Andi for review.
- Cyrill -
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
[Cyrill Gorcunov - Wed, Nov 26, 2008 at 11:10:54PM +0300]
...
| | | | The uncommon path is supposed to be out of line. I don't think | | the CPU will like that. | | | | -Andi | | | | > rff_action: | | | | -- | | [email protected] | | | | Aha! Thanks Andi for review. | | - Cyrill -
And if I would be tracing all and everywhere would it be common then? :)
- Cyrill -
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
In reply to Cyrill GorcunovQuoted 25 lines · Cyrill Gorcunov
[Andi Kleen - Wed, Nov 26, 2008 at 09:04:33PM +0100]
| [email protected] writes: | > --- a/arch/x86/kernel/entry_64.S | > +++ b/arch/x86/kernel/entry_64.S | > @@ -379,7 +379,10 @@ ENTRY(ret_from_fork) | > GET_THREAD_INFO(%rcx) | > testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT),TI_flags(%rcx) | > CFI_REMEMBER_STATE | > - jnz rff_trace | > + jz rff_action | > + movq %rsp,%rdi | > + call syscall_trace_leave | > + GET_THREAD_INFO(%rcx) | | The uncommon path is supposed to be out of line. I don't think | the CPU will like that. | | -Andi | | > rff_action: | | -- | [email protected] |
Quoted 1 line
Aha! Thanks Andi for review.
Unfortunately that review got you off the right track ...
The principle you applied was good: entry_64.S is murky, and we want to simplify the current overcomplicated assembly code flow spaghetti there.
Note that if you take a closer look at rff_trace/rff_action ret_from_fork code here, you'll realize that it does all the wrong
things: for example it checks the TIF flag - while later on jumping back to the ret-from-syscall path - duplicating the check needlessly.
But it gets worse than that: checking for _TIF_SYSCALL_TRACE is completely unnecessary here because we clear that flag for every freshly forked task. So the whole "tracing" code here, for which there is this "out of line jump optimization" is in reality completely dead code in the ptrace case...
Could you test something like the patch attached below, which cleans up this code and applies the code reduction and speedup? Warning:
completely untested! Please check that things like strace -f and gdb attaching to forked tasks still works fine. (it should by all means)
Thanks,
Ingo
--- arch/x86/kernel/entry_64.S | 31 ++++++++++++++++--------------- 1 file changed, 16 insertions(+), 15 deletions(-)
arch/x86/kernel/entry_64.S +16 -15
@@ -366,34 +366,35 @@ ENTRY(save_paranoid) END(save_paranoid) /* - * A newly forked process directly context switches into this. + * A newly forked process directly context switches into this address. + * + * rdi: prev task we switched from */ -/* rdi: prev */ ENTRY(ret_from_fork) DEFAULT_FRAME + push kernel_eflags(%rip) CFI_ADJUST_CFA_OFFSET 8 - popf # reset kernel eflags + popf # reset kernel eflags CFI_ADJUST_CFA_OFFSET -8 - call schedule_tail + + call schedule_tail # rdi: 'prev' task parameter + GET_THREAD_INFO(%rcx) - testl $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT),TI_flags(%rcx) + CFI_REMEMBER_STATE - jnz rff_trace -rff_action: RESTORE_REST - testl $3,CS-ARGOFFSET(%rsp) # from kernel_thread? + + testl $3, CS-ARGOFFSET(%rsp) # from kernel_thread? je int_ret_from_sys_call - testl $_TIF_IA32,TI_flags(%rcx) + + testl $_TIF_IA32, TI_flags(%rcx) # 32-bit compat task needs IRET jnz int_ret_from_sys_call + RESTORE_TOP_OF_STACK %rdi, -ARGOFFSET - jmp ret_from_sys_call + jmp ret_from_sys_call # go to the SYSRET fastpath + CFI_RESTORE_STATE -rff_trace: - movq %rsp,%rdi - call syscall_trace_leave - GET_THREAD_INFO(%rcx) - jmp rff_action CFI_ENDPROC END(ret_from_fork)
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
Quoted 2 lines
But it gets worse than that: checking for
_TIF_SYSCALL_TRACEis completely unnecessary here because we clear that flag for every
That's true. I found your 2005 changeset which did that, although it seems to have been written before meaningful changelogs became en vogue so it's unclear why. But assuming strace/gdb always reattach it's probably ok.
But _TIF_SYSCALL_AUDIT for which it also checks is not always cleared so it's not completely dead.
That said in theory the ret_from_sys checks should handle those too, so it might be unnecessary (I don't remember why I added the additional check here). Still removing it would change the order
of signal checks versus trace checks and since signal handling is always hairy one would need to be very careful when changing it to not break anything.
-Andi
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
Quoted 4 lines · Andi Kleen
Quoted 2 lines
But it gets worse than that: checking for
_TIF_SYSCALL_TRACEis completely unnecessary here because we clear that flag for everyThat's true. I found your 2005 changeset which did that, [...]
that's irrelevant, because all the necessary TIF_ flag processing is already done in ret_from_sys_call.
The unnecessary TIF_SYSCALL_TRACE code in the 64-bit ret_from_fork was apparently added by you in 2002:
+ENTRY(ret_from_fork)
+ movq %rbx, %rdi + call schedule_tail + GET_THREAD_INFO(%rcx) + bt $TIF_SYSCALL_TRACE,threadinfo_flags(%rcx) + jc rff_trace
when you added it to x86_64 via the changeset quoted below.
at that time the 32-bit code (from which you copied the x86_64 tree, creating the whole split x86 trees approach) had this simple and straightforward ret_from_fork implementation:
ENTRY(ret_from_fork)
#if CONFIG_SMP
pushl %ebx
call SYMBOL_NAME(schedule_tail)
addl $4, %esp
#endif
GET_THREAD_INFO(%ebx)
jmp syscall_exitso by all means the rff_trace/rff_action hack came from you.
Ingo
------------------>
Quoted 1 line
From
0457d99a336be658cea1a5bdb689de5adb3b382dMon Sep 17 00:00:00 2001
From: Andi Kleen <[email protected]>
Date: Tue, 12 Feb 2002 20:17:35 -0800
Subject: [PATCH] [PATCH] x86_64 merge: arch + asm
This adds the x86_64 arch and asm directories and a Documentation/x86_64.
It took a bit longer because I first had to make preemption and thread_info work and also found some other bugs while doing this. The port has been tested for a long time on UP.
I'm not sure what I should describe. A lot is based on i386 with a lot of cleanups. I wrote a paper about it for last year's OLS that describes most of the changes (ftp://ftp.firstfloor.org/pub/ak/x86_64.ps.gz). It is a bit outdated now, but should give a good overview.
It currently has a completely cut'n'pasted from others+hacked 32bit emulation. I hope to clean that up in the future by merging the generic core of this with other 64bit archs.
Thanks,
-Andi
--- Documentation/x86_64/mm.txt | 148 ++ arch/x86_64/Config.help | 531 +++++ arch/x86_64/Makefile | 130 ++ arch/x86_64/boot/Makefile | 92 + arch/x86_64/boot/bootsect.S | 416 ++++ arch/x86_64/boot/compressed/Makefile | 43 + arch/x86_64/boot/compressed/head.S | 142 ++ arch/x86_64/boot/compressed/misc.c | 431 ++++ arch/x86_64/boot/compressed/miscsetup.h | 39 + arch/x86_64/boot/install.sh | 39 + arch/x86_64/boot/setup.S | 955 ++++++++ arch/x86_64/boot/tools/build.c | 189 ++ arch/x86_64/boot/video.S | 1934 ++++++++++++++++ arch/x86_64/config.in | 240 ++ arch/x86_64/defconfig | 568 +++++ arch/x86_64/ia32/Makefile | 18 + arch/x86_64/ia32/ia32_binfmt.c | 165 ++ arch/x86_64/ia32/ia32_ioctl.c | 3843 +++++++++++++++++++++++++++++++ arch/x86_64/ia32/ia32_signal.c | 489 ++++ arch/x86_64/ia32/ia32entry.S | 365 +++ arch/x86_64/ia32/ptrace32.c | 312 +++ arch/x86_64/ia32/socket32.c | 686 ++++++ arch/x86_64/ia32/sys_ia32.c | 2788 ++++++++++++++++++++++ arch/x86_64/kernel/Makefile | 38 + arch/x86_64/kernel/apic.c | 1160 ++++++++++ arch/x86_64/kernel/bluesmoke.c | 174 ++ arch/x86_64/kernel/cpuid.c | 177 ++ arch/x86_64/kernel/early_printk.c | 77 + arch/x86_64/kernel/entry.S | 640 +++++ arch/x86_64/kernel/head.S | 342 +++ arch/x86_64/kernel/head64.c | 85 + arch/x86_64/kernel/i387.c | 439 ++++ arch/x86_64/kernel/i8259.c | 485 ++++ arch/x86_64/kernel/init_task.c | 42 + arch/x86_64/kernel/io_apic.c | 1617 +++++++++++++ arch/x86_64/kernel/ioport.c | 114 + arch/x86_64/kernel/irq.c | 1198 ++++++++++ arch/x86_64/kernel/ldt.c | 177 ++ arch/x86_64/kernel/mpparse.c | 670 ++++++ arch/x86_64/kernel/msr.c | 279 +++ arch/x86_64/kernel/mtrr.c | 2310 +++++++++++++++++++ arch/x86_64/kernel/nmi.c | 272 +++ arch/x86_64/kernel/pci-dma.c | 33 + arch/x86_64/kernel/pci-irq.c | 753 ++++++ arch/x86_64/kernel/pci-pc.c | 438 ++++ arch/x86_64/kernel/pci-x86_64.c | 384 +++ arch/x86_64/kernel/pci-x86_64.h | 72 + arch/x86_64/kernel/process.c | 756 ++++++ arch/x86_64/kernel/ptrace.c | 435 ++++ arch/x86_64/kernel/semaphore.c | 225 ++ arch/x86_64/kernel/setup.c | 1128 +++++++++ arch/x86_64/kernel/setup64.c | 142 ++ arch/x86_64/kernel/signal.c | 591 +++++ arch/x86_64/kernel/smp.c | 584 +++++ arch/x86_64/kernel/smpboot.c | 1023 ++++++++ arch/x86_64/kernel/sys_x86_64.c | 113 + arch/x86_64/kernel/syscall.c | 25 + arch/x86_64/kernel/time.c | 494 ++++ arch/x86_64/kernel/trampoline.S | 72 + arch/x86_64/kernel/traps.c | 771 +++++++ arch/x86_64/kernel/vsyscall.c | 190 ++ arch/x86_64/kernel/x8664_ksyms.c | 162 ++ arch/x86_64/lib/Makefile | 17 + arch/x86_64/lib/checksum_copy.S | 142 ++ arch/x86_64/lib/dec_and_lock.c | 40 + arch/x86_64/lib/delay.c | 45 + arch/x86_64/lib/generic-checksum.c | 124 + arch/x86_64/lib/getuser.S | 90 + arch/x86_64/lib/iodebug.c | 11 + arch/x86_64/lib/mmx.c | 377 +++ arch/x86_64/lib/old-checksum.c | 33 + arch/x86_64/lib/putuser.S | 88 + arch/x86_64/lib/rwsem_thunk.S | 27 + arch/x86_64/lib/usercopy.c | 147 ++ arch/x86_64/mm/Makefile | 13 + arch/x86_64/mm/extable.c | 62 + arch/x86_64/mm/fault.c | 324 +++ arch/x86_64/mm/init.c | 387 ++++ arch/x86_64/mm/ioremap.c | 163 ++ arch/x86_64/tools/Makefile | 31 + arch/x86_64/tools/offset.c | 70 + arch/x86_64/tools/offset.sed | 7 + arch/x86_64/vmlinux.lds | 112 + include/asm-x86_64/a.out.h | 30 + include/asm-x86_64/apic.h | 103 + include/asm-x86_64/apicdef.h | 363 +++ include/asm-x86_64/atomic.h | 206 ++ include/asm-x86_64/bitops.h | 465 ++++ include/asm-x86_64/boot.h | 15 + include/asm-x86_64/bootsetup.h | 34 + include/asm-x86_64/bugs.h | 42 + include/asm-x86_64/byteorder.h | 32 + include/asm-x86_64/cache.h | 13 + include/asm-x86_64/calling.h | 100 + include/asm-x86_64/checksum.h | 158 ++ include/asm-x86_64/cpufeature.h | 73 + include/asm-x86_64/current.h | 29 + include/asm-x86_64/debugreg.h | 65 + include/asm-x86_64/delay.h | 20 + include/asm-x86_64/desc.h | 187 ++ include/asm-x86_64/div64.h | 14 + include/asm-x86_64/dma.h | 298 +++ include/asm-x86_64/e820.h | 40 + include/asm-x86_64/elf.h | 120 + include/asm-x86_64/errno.h | 132 ++ include/asm-x86_64/fcntl.h | 79 + include/asm-x86_64/fixmap.h | 105 + include/asm-x86_64/floppy.h | 286 +++ include/asm-x86_64/hardirq.h | 93 + include/asm-x86_64/hdreg.h | 12 + include/asm-x86_64/hw_irq.h | 210 ++ include/asm-x86_64/i387.h | 87 + include/asm-x86_64/ia32.h | 274 +++ include/asm-x86_64/ia32_unistd.h | 251 ++ include/asm-x86_64/ide.h | 128 + include/asm-x86_64/init.h | 1 + include/asm-x86_64/io.h | 270 +++ include/asm-x86_64/io_apic.h | 148 ++ include/asm-x86_64/ioctl.h | 75 + include/asm-x86_64/ioctls.h | 82 + include/asm-x86_64/ipc.h | 32 + include/asm-x86_64/ipcbuf.h | 29 + include/asm-x86_64/irq.h | 35 + include/asm-x86_64/kdebug.h | 23 + include/asm-x86_64/keyboard.h | 71 + include/asm-x86_64/kmap_types.h | 14 + include/asm-x86_64/ldt.h | 37 + include/asm-x86_64/linux_logo.h | 29 + include/asm-x86_64/locks.h | 135 ++ include/asm-x86_64/mc146818rtc.h | 29 + include/asm-x86_64/mman.h | 39 + include/asm-x86_64/mmu.h | 13 + include/asm-x86_64/mmu_context.h | 95 + include/asm-x86_64/mmx.h | 14 + include/asm-x86_64/module.h | 12 + include/asm-x86_64/mpspec.h | 188 ++ include/asm-x86_64/msgbuf.h | 27 + include/asm-x86_64/msr.h | 140 ++ include/asm-x86_64/mtrr.h | 127 + include/asm-x86_64/namei.h | 17 + include/asm-x86_64/page.h | 118 + include/asm-x86_64/param.h | 24 + include/asm-x86_64/parport.h | 18 + include/asm-x86_64/pci.h | 273 +++ include/asm-x86_64/pda.h | 79 + include/asm-x86_64/pgalloc.h | 258 +++ include/asm-x86_64/pgtable.h | 399 ++++ include/asm-x86_64/poll.h | 25 + include/asm-x86_64/posix_types.h | 116 + include/asm-x86_64/prctl.h | 10 + include/asm-x86_64/processor.h | 463 ++++ include/asm-x86_64/ptrace.h | 114 + include/asm-x86_64/resource.h | 47 + include/asm-x86_64/rwlock.h | 84 + include/asm-x86_64/rwsem.h | 217 ++ include/asm-x86_64/scatterlist.h | 13 + include/asm-x86_64/segment.h | 19 + include/asm-x86_64/semaphore.h | 216 ++ include/asm-x86_64/sembuf.h | 25 + include/asm-x86_64/serial.h | 133 ++ include/asm-x86_64/setup.h | 10 + include/asm-x86_64/shmbuf.h | 38 + include/asm-x86_64/shmparam.h | 6 + include/asm-x86_64/sigcontext.h | 97 + include/asm-x86_64/siginfo.h | 232 ++ include/asm-x86_64/signal.h | 205 ++ include/asm-x86_64/smp.h | 98 + include/asm-x86_64/smplock.h | 95 + include/asm-x86_64/socket.h | 64 + include/asm-x86_64/socket32.h | 70 + include/asm-x86_64/sockios.h | 12 + include/asm-x86_64/softirq.h | 34 + include/asm-x86_64/spinlock.h | 181 ++ include/asm-x86_64/stat.h | 27 + include/asm-x86_64/statfs.h | 25 + include/asm-x86_64/string.h | 38 + include/asm-x86_64/system.h | 283 +++ include/asm-x86_64/termbits.h | 172 ++ include/asm-x86_64/termios.h | 106 + include/asm-x86_64/thread_info.h | 116 + include/asm-x86_64/timex.h | 51 + include/asm-x86_64/tlb.h | 1 + include/asm-x86_64/types.h | 47 + include/asm-x86_64/uaccess.h | 373 +++ include/asm-x86_64/ucontext.h | 12 + include/asm-x86_64/unaligned.h | 37 + include/asm-x86_64/unistd.h | 653 ++++++ include/asm-x86_64/user.h | 113 + include/asm-x86_64/user32.h | 58 + include/asm-x86_64/vga.h | 20 + include/asm-x86_64/vsyscall.h | 48 + include/asm-x86_64/xor.h | 859 +++++++ 192 files changed, 48138 insertions(+), 0 deletions(-)
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
On Fri, Nov 28, 2008 at 02:47:54PM +0100, Ingo Molnar wrote:
Hi Ingo,
Quoted 10 lines · Andi Kleen
Quoted 5 lines
Quoted 3 lines
But it gets worse than that: checking for
_TIF_SYSCALL_TRACEis completely unnecessary here because we clear that flag for everyThat's true. I found your 2005 changeset which did that, [...]
that's irrelevant, because all the necessary
TIF_flag processing is already done inret_from_sys_call.
Yes, that's true (as I wrote in my email), but it's just not dead because the audit code will take it (or used to handle both audit and fork before 2005), just unnecessary because the later code checks it as well@). Ok it's only a very subtle difference given.
It would have been nice if you had pruned at least the fork case when you changed fork.c to always clear it, but that can be also done now.
Quoted 3 lines
The unnecessary
TIF_SYSCALL_TRACEcode in the 64-bitret_from_forkwas apparently added by you in 2002:
Yes, to be honest I don't remember why I did it this way. Most likely the additional check was needed in some earlier iteration and then not pruned away when it became unnecessary. I remember I had some ordering problems with strace very early on it might have been related to that. Or perhaps I just goofed up. In the current form it's certainly not great code.
Anyways I have no objections to removing it now, just one has to be careful that the changed processing order (signals vs trace) -- which is likely user visible -- won't break anything.
-Andi
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
In reply to Ingo Molnar[Ingo Molnar - Thu, Nov 27, 2008 at 02:41:21PM +0100]
...
| | Could you test something like the patch attached below, which cleans | up this code and applies the code reduction and speedup? Warning: | completely untested! Please check that things like strace -f and gdb | attaching to forked tasks still works fine. (it should by all means) |
...
Thanks for explanation Ingo! Will try to test this but I will require some time.
- Cyrill -
Re: [PATCH 2/5] x86: ret_from_fork - get rid of jump back
In reply to Ingo Molnar[Ingo Molnar - Thu, Nov 27, 2008 at 02:41:21PM +0100]
| | * Cyrill Gorcunov <[email protected]> wrote: | ... | | Could you test something like the patch attached below, which cleans | up this code and applies the code reduction and speedup? Warning: | completely untested! Please check that things like strace -f and gdb | attaching to forked tasks still works fine. (it should by all means) | | Thanks, | | Ingo
Ingo, I found 2.6.26 test machine where I've patched the entry_64.S in a manner you pointed. So here is a test program and strace output. Seems all works fine. Didn't check "audit" code which I simply don't know how to be done.
- Cyrill -
---
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <sys/types.h>
#include <unistd.h>
int main(int argc, char *argv[])
{
int pid;
pid = fork();
if (pid < 0) {
fprintf(stderr, "Fork failed!\n");
exit(-1);
}
else if (pid == 0) {
printf("I am the child, return from fork=%d\n", pid);
} else {
printf("I am the parent, return from fork, child pid=%d\n", pid);
printf("Parent exiting!\n");
exit(0);
}
}---
5897 execve("./main", ["./main"], [/* 18 vars */]) = 0
5897 brk(0) = 0x2514000
5897 mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa0e517f000
5897 access("/etc/ld.so.nohwcap", F_OK) = -1 ENOENT (No such file or directory)
5897 mmap(NULL, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa0e517d000
5897 access("/etc/ld.so.preload", R_OK) = -1 ENOENT (No such file or directory)
5897 open("/etc/ld.so.cache", O_RDONLY) = 3
5897 fstat(3, {st_mode=S_IFREG|0644, st_size=48530, ...}) = 0
5897 mmap(NULL, 48530, PROT_READ, MAP_PRIVATE, 3, 0) = 0x7fa0e5171000
5897 close(3) = 0
5897 access("/etc/ld.so.nohwcap", F_OK) = -1 ENOENT (No such file or directory)
5897 open("/lib/libc.so.6", O_RDONLY) = 3
5897 read(3, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\220\345"..., 832) = 832
5897 fstat(3, {st_mode=S_IFREG|0755, st_size=1502520, ...}) = 0
5897 mmap(NULL, 3609304, PROT_READ|PROT_EXEC, MAP_PRIVATE|MAP_DENYWRITE, 3, 0) = 0x7fa0e4bf0000
5897 mprotect(0x7fa0e4d59000, 2093056, PROT_NONE) = 0
5897 mmap(0x7fa0e4f58000, 20480, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_DENYWRITE, 3, 0x168000) = 0x7fa0e4f58000
5897 mmap(0x7fa0e4f5d000, 17112, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|MAP_ANONYMOUS, -1, 0) = 0x7fa0e4f5d000
5897 close(3) = 0
5897 mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa0e5170000
5897 mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa0e516f000
5897 arch_prctl(ARCH_SET_FS, 0x7fa0e516f6e0) = 0
5897 mprotect(0x7fa0e4f58000, 16384, PROT_READ) = 0
5897 mprotect(0x7fa0e5180000, 4096, PROT_READ) = 0
5897 munmap(0x7fa0e5171000, 48530) = 0
5897 clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7fa0e516f770) = 5898
5897 fstat(1, {st_mode=S_IFCHR|0600, st_rdev=makedev(5, 1), ...}) = 0
5898 fstat(1, <unfinished ...>
5897 ioctl(1, SNDCTL_TMR_TIMEBASE or TCGETS <unfinished ...>
5898 <... fstat resumed> {st_mode=S_IFCHR|0600, st_rdev=makedev(5, 1), ...}) = 0
5897 <... ioctl resumed> , {B38400 opost isig icanon echo ...}) = 0
5897 mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0 <unfinished ...>
5898 ioctl(1, SNDCTL_TMR_TIMEBASE or TCGETS <unfinished ...>
5897 <... mmap resumed> ) = 0x7fa0e517c000
5898 <... ioctl resumed> , {B38400 opost isig icanon echo ...}) = 0
5897 write(1, "I am the parent, return from for"..., 50 <unfinished ...>
5898 mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0 <unfinished ...>
5897 <... write resumed> ) = 50
5898 <... mmap resumed> ) = 0x7fa0e517c000
5897 write(1, "Parent exiting!\n", 16 <unfinished ...>
5898 write(1, "I am the child, return from fork"..., 35 <unfinished ...>
5897 <... write resumed> ) = 16
5898 <... write resumed> ) = 35
5897 exit_group(0) = ?
5898 exit_group(0) = ?---
Re: [PATCH 1/5] x86: entry_64.S - use ENTRY to define child_rip
In reply to [email protected]Quoted 15 lines · On Wed, 26 Nov 2008 22:17:00 +0300, [email protected] said:
From: Cyrill Gorcunov <[email protected]>
child_ripis called not by its name but indirectly rather so make it global and aligned.Signed-off-by Cyrill Gorcunov <[email protected]>--- arch/x86/kernel/entry_64.S | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)arch/x86/kernel/entry_64.S +0
-0
arch/x86/kernel/entry_64.S ENTRY(kernel_thread) · line 1328
Quoted 8 lines
arch/x86/kernel/entry_64.S +1
-1@@ -1322,7 +1322,7 @@ ENTRY(kernel_thread) CFI_ENDPROC END(kernel_thread) -child_rip: +ENTRY(child_rip) pushq $0 # fake return address CFI_STARTPROC /*
Quoted 2 lines
1.6.0.4.603.gbc9c0
Signature · 6 lines
Alexander van Heukelum [email protected] -- http://www.fastmail.fm - Accessible with your email software or over the web
Ingo Molnar
AppliedRe: [PATCH 1/5] x86: entry_64.S - use ENTRY to define child_rip
Quoted 20 lines · Alexander van Heukelum
Quoted 18 lines · On Wed, 26 Nov 2008 22:17:00 +0300, [email protected] said:
From: Cyrill Gorcunov <[email protected]>
child_ripis called not by its name but indirectly rather so make it global and aligned.Signed-off-by Cyrill Gorcunov <[email protected]>--- arch/x86/kernel/entry_64.S | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)arch/x86/kernel/entry_64.S +0
-0@@ -1322,7 +1322,7 @@ ENTRY(kernel_thread) CFI_ENDPROC END(kernel_thread)
Quoted 7 lines
arch/x86/kernel/entry_64.S ENTRY(kernel_thread) · line 1328
Quoted 5 lines
arch/x86/kernel/entry_64.S +1
-1-child_rip: +ENTRY(child_rip) pushq $0 # fake return address CFI_STARTPROC /*
applied to tip/x86/irq, thanks!
Ingo