Skip to contentSkip to conversation

Linux 2.6.9-ac16

DISCUSSION Chuck Ebbert 1 participant 1 message Dec 21, 2004

lore
Share & tools
lore
#0 lore raw
Quoted 8 lines · Chris Friesen
Quoted 2 lines · Chuck Ebbert

Nobody has found an answer for the freezes, which persist even in the latest
2.6.10-rc but there's a vm_writeout throttling patch in -ac that I haven't tried.

Heh. Figures. Those freezes are what bothers me most, since I've already got a patch that protects critical processes from being OOM-killed as long as they're sane.

I backported this patch to 2.6.9 but haven't tested it yet. It requires the
'spurious oomkill' patch I posted earlier in this thread. Early reports are that it stops the freezes during heavy paging.

Also appended is a newer patch from Andrew that may help as well. It's obviously
correct but I haven't done anything with it yet so it may not apply to 2.6.9.

===================================================================================

# mm_swap_token_timeout.patch
#
#       Disable token-based thrashing control when token
#       timeout is zero and make zero the default value.
#
#       NOTE: On 2.6.9 there is no way to change the
#       token timeout at runtime like there is on 2.6.10.
#       You must edit mm/thrash.c instead.
#
#       Patch by Con Kolivas for 2.6.10-rc3 20 Dec 2004
#       Backported to 2.6.9 by Chuck Ebbert
#
--- 2.6.9.1/mm/rmap.c
+++ 2.6.9.2/mm/rmap.c
@@ -395,6 +395,9 @@ int page_referenced(struct page *page, i
 {
        int referenced = 0;

+ if (!SWAP_TOKEN_TIMEOUT)
+ ignore_token = 1;
+

        if (page_test_and_clear_young(page))
                referenced++;
mm/thrash.c +4 -1
@@ -19,7 +19,10 @@ unsigned long swap_token_check;
 struct mm_struct * swap_token_mm = &init_mm;
 
 #define SWAP_TOKEN_CHECK_INTERVAL (HZ * 2)
-#define SWAP_TOKEN_TIMEOUT (HZ * 300)
+#define SWAP_TOKEN_TIMEOUT     0
+/*
+ * Currently disabled; Needs further code to work at HZ * 300.
+ */
 
 /*
  * Take the token away if the process had no page faults

===================================================================================
# We haven't been incrementing local variable total_scanned since the

# scan_control stuff went in.  That broke kswapd throttling.
# 
# Signed-off-by: Andrew Morton <[email protected]>
# ---
# 
#  25-akpm/mm/vmscan.c |    1 +
#  1 files changed, 1 insertion(+)
# 
# diff -puN mm/vmscan.c~vmscan-total_scanned-fix mm/vmscan.c
--- 25/mm/vmscan.c~vmscan-total_scanned-fix     2004-12-20 12:47:25.855643408 -0800
+++ 25-akpm/mm/vmscan.c 2004-12-20 12:47:25.860642648 -0800
@@ -1063,6 +1063,7 @@ scan:
                        shrink_slab(sc.nr_scanned, GFP_KERNEL, lru_pages);
                        sc.nr_reclaimed += reclaim_state->reclaimed_slab;
                        total_reclaimed += sc.nr_reclaimed;
+                       total_scanned += sc.nr_scanned;
                        if (zone->all_unreclaimable)
                                continue;
                        if (zone->pages_scanned >= (zone->nr_active +
===================================================================================
--
Please take it as a sign of my infinite respect for you,
that I insist on you doing all the work.
                                        -- Rusty Russell