On Thu, Apr 22, 2021 at 03:38:53PM +0200, Martin Pieuchot wrote:
> Diff below remove the KERNEL_LOCK()/UNLOCK() dance from uvm_fault() for
> both amd64 and sparc64. That means the kernel lock will only be taken
> for lower faults and some amap/anon code will now run without it.
> I'd be interested to have this tested and see how much does that impact
> the build time of packages.
> We should be able to do the switch on an arch-by-arch basis. It's
> easier for me to develop & debug on these two architectures so I started
> with them. If you want to unlock another architecture and report back,
> I'd be glad.
Sorry for the delay on starting such a build, but one just finished for me
overnight. I don't know that it made much difference in time to build, but
it completed the full build just fine.
(It was 3 hours quicker, but it also built less packages, in particular
because all the ruby 2.6 packages are gone)
--Kurt
> Thanks,
> Martin
>
> Index: arch/amd64/amd64/trap.c
> ===================================================================
> RCS file: /cvs/src/sys/arch/amd64/amd64/trap.c,v
> retrieving revision 1.87
> diff -u -p -r1.87 trap.c
> --- arch/amd64/amd64/trap.c 22 Oct 2020 13:41:51 -0000 1.87
> +++ arch/amd64/amd64/trap.c 22 Apr 2021 13:06:54 -0000
> @@ -176,10 +176,7 @@ upageflttrap(struct trapframe *frame, ui
> union sigval sv;
> int signal, sicode, error;
>
> - KERNEL_LOCK();
> error = uvm_fault(&p->p_vmspace->vm_map, va, 0, access_type);
> - KERNEL_UNLOCK();
> -
> if (error == 0) {
> uvm_grow(p, va);
> return 1;
> @@ -261,9 +258,7 @@ kpageflttrap(struct trapframe *frame, ui
> if (curcpu()->ci_inatomic == 0 || map == kernel_map) {
> onfault = pcb->pcb_onfault;
> pcb->pcb_onfault = NULL;
> - KERNEL_LOCK();
> error = uvm_fault(map, va, 0, access_type);
> - KERNEL_UNLOCK();
> pcb->pcb_onfault = onfault;
>
> if (error == 0 && map != kernel_map)
> Index: arch/sparc64/sparc64/trap.c
> ===================================================================
> RCS file: /cvs/src/sys/arch/sparc64/sparc64/trap.c,v
> retrieving revision 1.108
> diff -u -p -r1.108 trap.c
> --- arch/sparc64/sparc64/trap.c 11 Mar 2021 11:17:00 -0000 1.108
> +++ arch/sparc64/sparc64/trap.c 22 Apr 2021 13:06:54 -0000
> @@ -773,10 +773,7 @@ data_access_fault(struct trapframe64 *tf
> if (!(addr & TLB_TAG_ACCESS_CTX)) {
> /* CTXT == NUCLEUS */
>
> - KERNEL_LOCK();
> error = uvm_fault(kernel_map, va, 0, access_type);
> - KERNEL_UNLOCK();
> -
> if (error == 0)
> return;
> goto kfault;
> @@ -792,9 +789,7 @@ data_access_fault(struct trapframe64 *tf
>
> onfault = (vaddr_t)p->p_addr->u_pcb.pcb_onfault;
> p->p_addr->u_pcb.pcb_onfault = NULL;
> - KERNEL_LOCK();
> error = uvm_fault(&p->p_vmspace->vm_map, (vaddr_t)va, 0, access_type);
> - KERNEL_UNLOCK();
> p->p_addr->u_pcb.pcb_onfault = (void *)onfault;
>
> /*
> @@ -959,9 +954,7 @@ text_access_fault(struct trapframe64 *tf
> uvm_map_inentry_sp, p->p_vmspace->vm_map.sserial))
> goto out;
>
> - KERNEL_LOCK();
> error = uvm_fault(&p->p_vmspace->vm_map, va, 0, access_type);
> - KERNEL_UNLOCK();
>
> /*
> * If this was a stack access we keep track of the maximum
> @@ -1055,9 +1048,7 @@ text_access_error(struct trapframe64 *tf
> uvm_map_inentry_sp, p->p_vmspace->vm_map.sserial))
> goto out;
>
> - KERNEL_LOCK();
> error = uvm_fault(&p->p_vmspace->vm_map, va, 0, access_type);
> - KERNEL_UNLOCK();
>
> /*
> * If this was a stack access we keep track of the maximum
