search for: __per_cpu_offset

...gt;> +"__raw_callee_save___kvm_vcpu_is_preempted:" > >> +FRAME_BEGIN > >> +"push %rdi;" > >> +"push %rdx;" > >> +"movslq %edi, %rdi;" > >> +"movq $steal_time+16, %rax;" > >> +"movq __per_cpu_offset(,%rdi,8), %rdx;" > >> +"cmpb $0, (%rdx,%rax);" Could we not put the $steal_time+16 displacement as an immediate in the cmpb and save a whole register here? That way we'd end up with something like: asm(" push %rdi; movslq %edi, %rdi; movq __per_cpu_offset(,%rd...

...gt;> +"__raw_callee_save___kvm_vcpu_is_preempted:" > >> +FRAME_BEGIN > >> +"push %rdi;" > >> +"push %rdx;" > >> +"movslq %edi, %rdi;" > >> +"movq $steal_time+16, %rax;" > >> +"movq __per_cpu_offset(,%rdi,8), %rdx;" > >> +"cmpb $0, (%rdx,%rax);" Could we not put the $steal_time+16 displacement as an immediate in the cmpb and save a whole register here? That way we'd end up with something like: asm(" push %rdi; movslq %edi, %rdi; movq __per_cpu_offset(,%rd...

...Peter Zijlstra wrote: > >> On Mon, Feb 13, 2017 at 11:47:16AM +0100, Peter Zijlstra wrote: > >>> That way we'd end up with something like: > >>> > >>> asm(" > >>> push %rdi; > >>> movslq %edi, %rdi; > >>> movq __per_cpu_offset(,%rdi,8), %rax; > >>> cmpb $0, %[offset](%rax); > >>> setne %al; > >>> pop %rdi; > >>> " : : [offset] "i" (((unsigned long)&steal_time) + offsetof(struct steal_time, preempted))); > >>> > >>> And if we could...

...Peter Zijlstra wrote: > >> On Mon, Feb 13, 2017 at 11:47:16AM +0100, Peter Zijlstra wrote: > >>> That way we'd end up with something like: > >>> > >>> asm(" > >>> push %rdi; > >>> movslq %edi, %rdi; > >>> movq __per_cpu_offset(,%rdi,8), %rax; > >>> cmpb $0, %[offset](%rax); > >>> setne %al; > >>> pop %rdi; > >>> " : : [offset] "i" (((unsigned long)&steal_time) + offsetof(struct steal_time, preempted))); > >>> > >>> And if we could...

On 02/13/2017 05:53 AM, Peter Zijlstra wrote: > On Mon, Feb 13, 2017 at 11:47:16AM +0100, Peter Zijlstra wrote: >> That way we'd end up with something like: >> >> asm(" >> push %rdi; >> movslq %edi, %rdi; >> movq __per_cpu_offset(,%rdi,8), %rax; >> cmpb $0, %[offset](%rax); >> setne %al; >> pop %rdi; >> " : : [offset] "i" (((unsigned long)&steal_time) + offsetof(struct steal_time, preempted))); >> >> And if we could get rid of the sign extend on edi we could avoid all t...

On 02/13/2017 05:53 AM, Peter Zijlstra wrote: > On Mon, Feb 13, 2017 at 11:47:16AM +0100, Peter Zijlstra wrote: >> That way we'd end up with something like: >> >> asm(" >> push %rdi; >> movslq %edi, %rdi; >> movq __per_cpu_offset(,%rdi,8), %rax; >> cmpb $0, %[offset](%rax); >> setne %al; >> pop %rdi; >> " : : [offset] "i" (((unsigned long)&steal_time) + offsetof(struct steal_time, preempted))); >> >> And if we could get rid of the sign extend on edi we could avoid all t...

...ruary 13, 2017 2:53:43 AM PST, Peter Zijlstra <peterz at infradead.org> wrote: >On Mon, Feb 13, 2017 at 11:47:16AM +0100, Peter Zijlstra wrote: >> That way we'd end up with something like: >> >> asm(" >> push %rdi; >> movslq %edi, %rdi; >> movq __per_cpu_offset(,%rdi,8), %rax; >> cmpb $0, %[offset](%rax); >> setne %al; >> pop %rdi; >> " : : [offset] "i" (((unsigned long)&steal_time) + offsetof(struct >steal_time, preempted))); >> >> And if we could get rid of the sign extend on edi we could avoid...

...ruary 13, 2017 2:53:43 AM PST, Peter Zijlstra <peterz at infradead.org> wrote: >On Mon, Feb 13, 2017 at 11:47:16AM +0100, Peter Zijlstra wrote: >> That way we'd end up with something like: >> >> asm(" >> push %rdi; >> movslq %edi, %rdi; >> movq __per_cpu_offset(,%rdi,8), %rax; >> cmpb $0, %[offset](%rax); >> setne %al; >> pop %rdi; >> " : : [offset] "i" (((unsigned long)&steal_time) + offsetof(struct >steal_time, preempted))); >> >> And if we could get rid of the sign extend on edi we could avoid...

...r.c crash-6.0.8/xen_hyper.c --- crash-6.0.8.orig/xen_hyper.c 2012-07-05 15:47:09.000000000 +0200 +++ crash-6.0.8/xen_hyper.c 2012-07-05 15:50:19.000000000 +0200 @@ -64,7 +64,6 @@ xen_hyper_init(void) machdep->get_smp_cpus(); machdep->memory_size(); -#ifdef IA64 if (symbol_exists("__per_cpu_offset")) { xht->flags |= XEN_HYPER_SMP; if((xht->__per_cpu_offset = malloc(sizeof(ulong) * XEN_HYPER_MAX_CPUS())) == NULL) { @@ -76,7 +75,6 @@ xen_hyper_init(void) error(FATAL, "cannot read __per_cpu_offset.\n"); } } -#endif #if defined(X86) || defined(X86_64) if...

This patch series implements SMP boot for arch/arm, as far as getting all CPUs up and running the idle loop. Changes from v1: - moved barriers out of loop in udelay() - dropped broken GIC change in favour of explanatory comment - made the increment of ready_cpus atomic (I couldn''t move the increment to before signalling the next CPU because the PT switch has to happen between

Hi Andrew, This patch series adds to the end of the existing i386-gdt-cleanups patches: allow-per-cpu-variables-to-be-page-aligned.patch i386-gdt-cleanups-use-per-cpu-variables-for-gdt-pda.patch i386-gdt-cleanups-use-per-cpu-gdt-immediately-upon-boot.patch i386-gdt-cleanups-use-per-cpu-gdt-immediately-upon-boot-fix.patch i386-gdt-cleanups-clean-up-cpu_init.patch

Hi Andrew, This patch series adds to the end of the existing i386-gdt-cleanups patches: allow-per-cpu-variables-to-be-page-aligned.patch i386-gdt-cleanups-use-per-cpu-variables-for-gdt-pda.patch i386-gdt-cleanups-use-per-cpu-gdt-immediately-upon-boot.patch i386-gdt-cleanups-use-per-cpu-gdt-immediately-upon-boot-fix.patch i386-gdt-cleanups-clean-up-cpu_init.patch

...aw_callee_save___kvm_vcpu_is_preempted, @function;" > +"__raw_callee_save___kvm_vcpu_is_preempted:" > +FRAME_BEGIN > +"push %rdi;" > +"push %rdx;" > +"movslq %edi, %rdi;" > +"movq $steal_time+16, %rax;" > +"movq __per_cpu_offset(,%rdi,8), %rdx;" > +"cmpb $0, (%rdx,%rax);" > +"setne %al;" > +"pop %rdx;" > +"pop %rdi;" > +FRAME_END > +"ret;" > +".popsection"); > + > +#endif > + > /* > * Setup pv_lock_ops to exploit KV...

...aw_callee_save___kvm_vcpu_is_preempted, @function;" > +"__raw_callee_save___kvm_vcpu_is_preempted:" > +FRAME_BEGIN > +"push %rdi;" > +"push %rdx;" > +"movslq %edi, %rdi;" > +"movq $steal_time+16, %rax;" > +"movq __per_cpu_offset(,%rdi,8), %rdx;" > +"cmpb $0, (%rdx,%rax);" > +"setne %al;" > +"pop %rdx;" > +"pop %rdi;" > +FRAME_END > +"ret;" > +".popsection"); > + > +#endif > + > /* > * Setup pv_lock_ops to exploit KV...

Hi Andi, This is a series of patches based on your latest queue (as of the other day, at least). It includes: - the most recent patch to compute the appropriate amount of percpu space to allocate, using a separate reservation for modules where needed. - make the percpu sections page-aligned, so that percpu variables can be page aligned if needed (which is used by gdt_page) -

Hi Andi, This is a series of patches based on your latest queue (as of the other day, at least). It includes: - the most recent patch to compute the appropriate amount of percpu space to allocate, using a separate reservation for modules where needed. - make the percpu sections page-aligned, so that percpu variables can be page aligned if needed (which is used by gdt_page) -

...n/xen-asm_32.S b/arch/x86/xen/xen-asm_32.S > index f9643fc..33ca6e4 100644 > --- a/arch/x86/xen/xen-asm_32.S > +++ b/arch/x86/xen/xen-asm_32.S > @@ -89,11 +89,11 @@ ENTRY(xen_iret) > */ > #ifdef CONFIG_SMP > GET_THREAD_INFO(%eax) > - movl TI_cpu(%eax), %eax > - movl __per_cpu_offset(,%eax,4), %eax > - mov xen_vcpu(%eax), %eax > + movl %ss:TI_cpu(%eax), %eax > + movl %ss:__per_cpu_offset(,%eax,4), %eax > + mov %ss:xen_vcpu(%eax), %eax > #else > - movl xen_vcpu, %eax > + movl %ss:xen_vcpu, %eax > #endif > > /* check IF state we're restorin...

...n/xen-asm_32.S b/arch/x86/xen/xen-asm_32.S > index f9643fc..33ca6e4 100644 > --- a/arch/x86/xen/xen-asm_32.S > +++ b/arch/x86/xen/xen-asm_32.S > @@ -89,11 +89,11 @@ ENTRY(xen_iret) > */ > #ifdef CONFIG_SMP > GET_THREAD_INFO(%eax) > - movl TI_cpu(%eax), %eax > - movl __per_cpu_offset(,%eax,4), %eax > - mov xen_vcpu(%eax), %eax > + movl %ss:TI_cpu(%eax), %eax > + movl %ss:__per_cpu_offset(,%eax,4), %eax > + mov %ss:xen_vcpu(%eax), %eax > #else > - movl xen_vcpu, %eax > + movl %ss:xen_vcpu, %eax > #endif > > /* check IF state we're restorin...

On Mon, Feb 13, 2017 at 11:47:16AM +0100, Peter Zijlstra wrote: > That way we'd end up with something like: > > asm(" > push %rdi; > movslq %edi, %rdi; > movq __per_cpu_offset(,%rdi,8), %rax; > cmpb $0, %[offset](%rax); > setne %al; > pop %rdi; > " : : [offset] "i" (((unsigned long)&steal_time) + offsetof(struct steal_time, preempted))); > > And if we could get rid of the sign extend on edi we could avoid all the > push-pop nonsen...

...rote: > On 02/13/2017 05:53 AM, Peter Zijlstra wrote: >> On Mon, Feb 13, 2017 at 11:47:16AM +0100, Peter Zijlstra wrote: >>> That way we'd end up with something like: >>> >>> asm(" >>> push %rdi; >>> movslq %edi, %rdi; >>> movq __per_cpu_offset(,%rdi,8), %rax; >>> cmpb $0, %[offset](%rax); >>> setne %al; >>> pop %rdi; >>> " : : [offset] "i" (((unsigned long)&steal_time) + offsetof(struct steal_time, preempted))); >>> >>> And if we could get rid of the sign extend on...