4.4-stable review patch. If anyone has any objections, please let me know. ------------------ From: Mike Kravetz <mike.kravetz@xxxxxxxxxx> [ Upstream commit 654f4807624a657f364417c2a7454f0df9961734 ] When a TSB grows beyond its current capacity, a new TSB is allocated and copy_tsb is called to copy entries from the old TSB to the new. A hash shift based on page size is used to calculate the index of an entry in the TSB. copy_tsb has hard coded PAGE_SHIFT in these calculations. However, for huge page TSBs the value REAL_HPAGE_SHIFT should be used. As a result, when copy_tsb is called for a huge page TSB the entries are placed at the incorrect index in the newly allocated TSB. When doing hardware table walk, the MMU does not match these entries and we end up in the TSB miss handling code. This code will then create and write an entry to the correct index in the TSB. We take a performance hit for the table walk miss and recreation of these entries. Pass a new parameter to copy_tsb that is the page size shift to be used when copying the TSB. Suggested-by: Anthony Yznaga <anthony.yznaga@xxxxxxxxxx> Signed-off-by: Mike Kravetz <mike.kravetz@xxxxxxxxxx> Signed-off-by: David S. Miller <davem@xxxxxxxxxxxxx> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx> --- arch/sparc/kernel/tsb.S | 11 +++++++---- arch/sparc/mm/tsb.c | 7 +++++-- 2 files changed, 12 insertions(+), 6 deletions(-) --- a/arch/sparc/kernel/tsb.S +++ b/arch/sparc/kernel/tsb.S @@ -470,13 +470,16 @@ __tsb_context_switch: .type copy_tsb,#function copy_tsb: /* %o0=old_tsb_base, %o1=old_tsb_size * %o2=new_tsb_base, %o3=new_tsb_size + * %o4=page_size_shift */ sethi %uhi(TSB_PASS_BITS), %g7 srlx %o3, 4, %o3 - add %o0, %o1, %g1 /* end of old tsb */ + add %o0, %o1, %o1 /* end of old tsb */ sllx %g7, 32, %g7 sub %o3, 1, %o3 /* %o3 == new tsb hash mask */ + mov %o4, %g1 /* page_size_shift */ + 661: prefetcha [%o0] ASI_N, #one_read .section .tsb_phys_patch, "ax" .word 661b @@ -501,9 +504,9 @@ copy_tsb: /* %o0=old_tsb_base, %o1=old_ /* This can definitely be computed faster... */ srlx %o0, 4, %o5 /* Build index */ and %o5, 511, %o5 /* Mask index */ - sllx %o5, PAGE_SHIFT, %o5 /* Put into vaddr position */ + sllx %o5, %g1, %o5 /* Put into vaddr position */ or %o4, %o5, %o4 /* Full VADDR. */ - srlx %o4, PAGE_SHIFT, %o4 /* Shift down to create index */ + srlx %o4, %g1, %o4 /* Shift down to create index */ and %o4, %o3, %o4 /* Mask with new_tsb_nents-1 */ sllx %o4, 4, %o4 /* Shift back up into tsb ent offset */ TSB_STORE(%o2 + %o4, %g2) /* Store TAG */ @@ -511,7 +514,7 @@ copy_tsb: /* %o0=old_tsb_base, %o1=old_ TSB_STORE(%o2 + %o4, %g3) /* Store TTE */ 80: add %o0, 16, %o0 - cmp %o0, %g1 + cmp %o0, %o1 bne,pt %xcc, 90b nop --- a/arch/sparc/mm/tsb.c +++ b/arch/sparc/mm/tsb.c @@ -451,7 +451,8 @@ retry_tsb_alloc: extern void copy_tsb(unsigned long old_tsb_base, unsigned long old_tsb_size, unsigned long new_tsb_base, - unsigned long new_tsb_size); + unsigned long new_tsb_size, + unsigned long page_size_shift); unsigned long old_tsb_base = (unsigned long) old_tsb; unsigned long new_tsb_base = (unsigned long) new_tsb; @@ -459,7 +460,9 @@ retry_tsb_alloc: old_tsb_base = __pa(old_tsb_base); new_tsb_base = __pa(new_tsb_base); } - copy_tsb(old_tsb_base, old_size, new_tsb_base, new_size); + copy_tsb(old_tsb_base, old_size, new_tsb_base, new_size, + tsb_index == MM_TSB_BASE ? + PAGE_SHIFT : REAL_HPAGE_SHIFT); } mm->context.tsb_block[tsb_index].tsb = new_tsb;