From: Hugh Dickins <hugh@veritas.com>
The low-latency unmap_vmas patch silently moved the zap_bytes test after the
TLB finish and lockbreak and regather: why? That not only makes zap_bytes
redundant (might as well use ZAP_BLOCK_SIZE), it makes the unmap_vmas level
redundant too - it's all about saving TLB flushes when unmapping a series of
small vmas.
Move zap_bytes test back before the lockbreak, and delete the curious comment
that a small zap block size doesn't matter: it's true need_flush prevents TLB
flush when no page has been unmapped, but unmapping pages in small blocks
involves many more TLB flushes than in large blocks.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
---
25-akpm/mm/memory.c | 9 +++------
1 files changed, 3 insertions(+), 6 deletions(-)
diff -puN mm/memory.c~vmtrunc-restore-unmap_vmas-zap_bytes mm/memory.c
--- 25/mm/memory.c~vmtrunc-restore-unmap_vmas-zap_bytes 2004-11-17 20:46:38.297535408 -0800
+++ 25-akpm/mm/memory.c 2004-11-17 20:46:38.301534800 -0800
@@ -613,10 +613,6 @@ static void unmap_page_range(struct mmu_
}
#ifdef CONFIG_PREEMPT
-/*
- * It's not an issue to have a small zap block size - TLB flushes
- * only happen once normally, due to the tlb->need_flush optimization.
- */
# define ZAP_BLOCK_SIZE (8 * PAGE_SIZE)
#else
/* No preempt: go for improved straight-line efficiency */
@@ -695,6 +691,9 @@ int unmap_vmas(struct mmu_gather **tlbp,
start += block;
zap_bytes -= block;
+ if ((long)zap_bytes > 0)
+ continue;
+
if (!atomic) {
int fullmm = tlb_is_full_mm(*tlbp);
tlb_finish_mmu(*tlbp, tlb_start, start);
@@ -702,8 +701,6 @@ int unmap_vmas(struct mmu_gather **tlbp,
*tlbp = tlb_gather_mmu(mm, fullmm);
tlb_start_valid = 0;
}
- if ((long)zap_bytes > 0)
- continue;
zap_bytes = ZAP_BLOCK_SIZE;
}
}
_