diff options
author | David Vrabel <david.vrabel@citrix.com> | 2012-07-09 10:39:05 (GMT) |
---|---|---|
committer | Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> | 2012-07-19 19:51:43 (GMT) |
commit | d095d43e78dd811d5c02c25e207c3364019b5a77 (patch) | |
tree | 4d104a6b39487e3cddd3ba2490fe257ed3f5793d /arch/x86 | |
parent | 37a80bf560786d96c5e8370bff45d867e43fd5c3 (diff) | |
download | linux-d095d43e78dd811d5c02c25e207c3364019b5a77.tar.xz |
xen/mm: do direct hypercall in xen_set_pte() if batching is unavailable
In xen_set_pte() if batching is unavailable (because the caller is in
an interrupt context such as handling a page fault) it would fall back
to using native_set_pte() and trapping and emulating the PTE write.
On 32-bit guests this requires two traps for each PTE write (one for
each dword of the PTE). Instead, do one mmu_update hypercall
directly.
During construction of the initial page tables, continue to use
native_set_pte() because most of the PTEs being set are in writable
and unpinned pages (see phys_pmd_init() in arch/x86/mm/init_64.c) and
using a hypercall for this is very expensive.
This significantly improves page fault performance in 32-bit PV
guests.
lmbench3 test Before After Improvement
----------------------------------------------
lat_pagefault 3.18 us 2.32 us 27%
lat_proc fork 356 us 313.3 us 11%
Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Diffstat (limited to 'arch/x86')
-rw-r--r-- | arch/x86/xen/mmu.c | 30 |
1 files changed, 25 insertions, 5 deletions
diff --git a/arch/x86/xen/mmu.c b/arch/x86/xen/mmu.c index 3a73785..3f1783a 100644 --- a/arch/x86/xen/mmu.c +++ b/arch/x86/xen/mmu.c @@ -308,8 +308,20 @@ static bool xen_batched_set_pte(pte_t *ptep, pte_t pteval) static inline void __xen_set_pte(pte_t *ptep, pte_t pteval) { - if (!xen_batched_set_pte(ptep, pteval)) - native_set_pte(ptep, pteval); + if (!xen_batched_set_pte(ptep, pteval)) { + /* + * Could call native_set_pte() here and trap and + * emulate the PTE write but with 32-bit guests this + * needs two traps (one for each of the two 32-bit + * words in the PTE) so do one hypercall directly + * instead. + */ + struct mmu_update u; + + u.ptr = virt_to_machine(ptep).maddr | MMU_NORMAL_PT_UPDATE; + u.val = pte_val_ma(pteval); + HYPERVISOR_mmu_update(&u, 1, NULL, DOMID_SELF); + } } static void xen_set_pte(pte_t *ptep, pte_t pteval) @@ -1416,13 +1428,21 @@ static pte_t __init mask_rw_pte(pte_t *ptep, pte_t pte) } #endif /* CONFIG_X86_64 */ -/* Init-time set_pte while constructing initial pagetables, which - doesn't allow RO pagetable pages to be remapped RW */ +/* + * Init-time set_pte while constructing initial pagetables, which + * doesn't allow RO page table pages to be remapped RW. + * + * Many of these PTE updates are done on unpinned and writable pages + * and doing a hypercall for these is unnecessary and expensive. At + * this point it is not possible to tell if a page is pinned or not, + * so always write the PTE directly and rely on Xen trapping and + * emulating any updates as necessary. + */ static void __init xen_set_pte_init(pte_t *ptep, pte_t pte) { pte = mask_rw_pte(ptep, pte); - xen_set_pte(ptep, pte); + native_set_pte(ptep, pte); } static void pin_pagetable_pfn(unsigned cmd, unsigned long pfn) |