lk/linux-kernel-test
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

powerpc: Only demote individual slices rather than whole process

Browse Source 
At present, if we have a kernel with a 64kB page size, and some
process maps something that has to be mapped with 4kB pages (such as a
cache-inhibited mapping on POWER5+, or the eHCA infiniband queue-pair
pages), we change the process to use 4kB pages everywhere.  This hurts
the performance of HPC programs that access eHCA from userspace.

With this patch, the kernel will only demote the slice(s) containing
the eHCA or cache-inhibited mappings, leaving the remaining slices
able to use 64kB hardware pages.

This also changes the slice_get_unmapped_area code so that it is
willing to place a 64k-page mapping into (or across) a 4k-page slice
if there is no better alternative, i.e. if the program specified
MAP_FIXED or if there is not sufficient space available in slices that
are either empty or already have 64k-page mappings in them.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
This commit is contained in:
Paul Mackerras
2008-06-18 15:29:12 +10:00
parent e952e6c4d6
commit 3a8247cc2c
4 changed files with 168 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
arch/powerpc/mm/hash_utils_64.c
View File

@ -695,6 +695,28 @@ unsigned int hash_page_do_lazy_icache(unsigned int pp, pte_t pte, int trap)
return pp;
}
#ifdef CONFIG_PPC_MM_SLICES
unsigned int get_paca_psize(unsigned long addr)
{
unsigned long index, slices;
if (addr < SLICE_LOW_TOP) {
slices = get_paca()->context.low_slices_psize;
index = GET_LOW_SLICE_INDEX(addr);
} else {
slices = get_paca()->context.high_slices_psize;
index = GET_HIGH_SLICE_INDEX(addr);
}
return (slices >> (index * 4)) & 0xF;
}
#else
unsigned int get_paca_psize(unsigned long addr)
{
return get_paca()->context.user_psize;
}
#endif
/*
* Demote a segment to using 4k pages.
* For now this makes the whole process use 4k pages.
@ -702,13 +724,13 @@ unsigned int hash_page_do_lazy_icache(unsigned int pp, pte_t pte, int trap)
#ifdef CONFIG_PPC_64K_PAGES
void demote_segment_4k(struct mm_struct *mm, unsigned long addr)
{
if (mm->context.user_psize == MMU_PAGE_4K)
if (get_slice_psize(mm, addr) == MMU_PAGE_4K)
return;
slice_set_user_psize(mm, MMU_PAGE_4K);
slice_set_range_psize(mm, addr, 1, MMU_PAGE_4K);
#ifdef CONFIG_SPU_BASE
spu_flush_all_slbs(mm);
#endif
if (get_paca()->context.user_psize != MMU_PAGE_4K) {
if (get_paca_psize(addr) != MMU_PAGE_4K) {
get_paca()->context = mm->context;
slb_flush_and_rebolt();
}
@ -792,11 +814,7 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
DBG_LOW(" user region with no mm !\n");
return 1;
}
#ifdef CONFIG_PPC_MM_SLICES
psize = get_slice_psize(mm, ea);
#else
psize = mm->context.user_psize;
#endif
ssize = user_segment_size(ea);
vsid = get_vsid(mm->context.id, ea, ssize);
break;
@ -868,7 +886,7 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
/* Do actual hashing */
#ifdef CONFIG_PPC_64K_PAGES
/* If _PAGE_4K_PFN is set, make sure this is a 4k segment */
if (pte_val(*ptep) & _PAGE_4K_PFN) {
if ((pte_val(*ptep) & _PAGE_4K_PFN) && psize == MMU_PAGE_64K) {
demote_segment_4k(mm, ea);
psize = MMU_PAGE_4K;
}
@ -897,7 +915,7 @@ int hash_page(unsigned long ea, unsigned long access, unsigned long trap)
}
}
if (user_region) {
if (psize != get_paca()->context.user_psize) {
if (psize != get_paca_psize(ea)) {
get_paca()->context = mm->context;
slb_flush_and_rebolt();
}