Andrea,
I'm resending you Rik's patch but now with CC to the list, so people can
try it out.
Basically this patch returns success on do_try_to_free_pages() with any
small progress made, instead fix-mm-3 behaviour, which returns success
only if the number of requested pages are freed.
This gave better VM behaviour, but I still notice problems.
Now i'll try out to wait on dirty buffers at sync_page_buffers() (like 2.4
does), which may help.
--- vmscan.c.orig Fri Jan 3 03:25:04 1997
+++ vmscan.c Fri Jan 3 03:31:23 1997
@@ -380,6 +380,8 @@
static int do_try_to_free_pages(unsigned int gfp_mask)
{
int priority;
+ int ret = 0;
+ int swapcount;
int count = SWAP_CLUSTER_MAX;
lock_kernel();
@@ -390,6 +392,7 @@
priority = 6;
do {
while (shrink_mmap(priority, gfp_mask)) {
+ ret = 1;
if (!--count)
goto done;
}
@@ -397,15 +400,18 @@
/* Try to get rid of some shared memory pages.. */
if (gfp_mask & __GFP_IO) {
while (shm_swap(priority, gfp_mask)) {
+ ret = 1;
if (!--count)
goto done;
}
}
/* Then, try to page stuff out.. */
+ swapcount = count;
while (swap_out(priority, gfp_mask)) {
- if (!--count)
- goto done;
+ ret = 1;
+ if (!--swapcount)
+ break;
}
shrink_dcache_memory(priority, gfp_mask);
@@ -417,7 +423,7 @@
printk("VM: do_try_to_free_pages failed for %s...\n",
current->comm);
/* Return success if we freed a page. */
- return priority >= 0;
+ return ret;
}
/*
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Fri Jun 16 2000 - 20:45:04 EDT