OOPS: divide error while s2dsk (2.6.20-rc1-mm1)

DISCUSSION Jiri Slaby 5 participants 12 messages Dec 18, 2006 – Dec 19, 2006

Subsystems (3)

Memory management MGLRU MM reclaim

lore

Share & tools

lore

Jiri Slaby

author

2006-12-18 11:20 UTC #0 lore raw

Hi.

I got this oops while suspending:

[  309.366557] Disabling non-boot CPUs ...
[  309.386563] CPU 1 is now offline
[  309.387625] CPU1 is down
[  309.387704] Stopping tasks ... done.
[  310.030991] Shrinking memory... -<0>divide error: 0000 [#1]
[  310.456669] SMP
[  310.456814] last sysfs file:
/devices/pci0000:00/0000:00:1e.0/0000:02:08.0/eth0/statistics/collisions
[  310.456919] Modules linked in: eth1394 floppy ohci1394 ide_cd ieee1394 cdrom
[  310.457259] CPU:    0
[  310.457260] EIP:    0060:[<c0150c9a>]    Not tainted VLI
[  310.457261] EFLAGS: 00210246   (2.6.20-rc1-mm1 #207)
[  310.457478] EIP is at shrink_slab+0x9e/0x169
[  310.457548] eax: 00000000   ebx: 00000000   ecx: 00000000   edx: 00000000
[  310.457623] esi: 00000000   edi: c18fe500   ebp: f7b3fe3c   esp: f7b3fe08
[  310.457696] ds: 007b   es: 007b   fs: 00d8  gs: 0033  ss: 0068
[  310.457772] Process swsusp (pid: 3243, ti=f7b3e000 task=f756f030
task.ti=f7b3e000)
[  310.457845] Stack: c175d8a0 c175daa0 c175db00 00000000 00000000 c053cec0
000045ec 000000d0
[  310.458286]        00000000 00000000 00001179 00001179 00000000 f7b3fe94
c0151445 00000001
[  310.458723]        f7b3fe64 00001df1 00000002 00000001 00000001 00038000
00000c79 0000117b
[  310.459199] Call Trace:
[  310.459334]  [<c0103f1b>] show_trace_log_lvl+0x1a/0x30
[  310.459450]  [<c0103fd6>] show_stack_log_lvl+0xa5/0xca
[  310.459562]  [<c01041ce>] show_registers+0x1d3/0x2b8
[  310.459673]  [<c01043d4>] die+0x121/0x243
[  310.459781]  [<c010456c>] do_trap+0x76/0x9c
[  310.459892]  [<c0104bd8>] do_divide_error+0x94/0x9e
[  310.460001]  [<c038a7e4>] error_code+0x7c/0x84
[  310.460113]  [<c0151445>] shrink_all_memory+0x211/0x2eb
[  310.460225]  [<c01418c1>] swsusp_shrink_memory+0x187/0x196
[  310.460335]  [<c0141a07>] prepare_processes+0x35/0xc8
[  310.460446]  [<c0141cce>] pm_suspend_disk+0xd/0x16f
[  310.460558]  [<c0140c87>] enter_state+0x129/0x19b
[  310.460668]  [<c0140d9c>] state_store+0xa3/0xac
[  310.460777]  [<c0198ab0>] subsys_attr_store+0x20/0x25
[  310.460889]  [<c0198b9f>] sysfs_write_file+0x97/0xd8
[  310.460998]  [<c0165262>] vfs_write+0x8b/0x149
[  310.461108]  [<c01658cb>] sys_write+0x3d/0x64
[  310.461216]  [<c0102fe4>] syscall_call+0x7/0xb
[  310.461328]  =======================
[  310.461397] Code: 31 c0 ff 17 89 c3 8b 45 e4 31 d2 f7 77 0c f7 e3 89 45 d8 89
55 dc 89 d1 89 c6 31 d2 85 c9 74 09 89 c8 31 d2 f7 75 f0 89 c1 89 f0 <f7> 75 f0
89 ca 89 45 d8 89 55 dc 8b 45 d8 03 47 10 89 47 10 85
[  310.464079] EIP: [<c0150c9a>] shrink_slab+0x9e/0x169 SS:ESP 0068:f7b3fe08
[  310.464228]

swsusp script is something like this:
echo platform > /sys/power/disk
echo disk > /sys/power/state

regards,

Signature · 4 lines

http://www.fi.muni.cz/~xslaby/            Jiri Slaby
faculty of informatics, masaryk university, brno, cz
e-mail: jirislaby gmail com, gpg pubkey fingerprint:
B674 9967 0407 CE62 ACC8  22A0 32CC 55C3 39D4 7A7E

Rafael J. Wysocki

2006-12-18 15:44 UTC #1 lore raw

Hi,

Quoted 16 lines · Jiri Slaby

Hi.

I got this oops while suspending:

[  309.366557] Disabling non-boot CPUs ...
[  309.386563] CPU 1 is now offline
[  309.387625] CPU1 is down
[  309.387704] Stopping tasks ... done.
[  310.030991] Shrinking memory... -<0>divide error: 0000 [#1]
[  310.456669] SMP
[  310.456814] last sysfs file:
/devices/pci0000:00/0000:00:1e.0/0000:02:08.0/eth0/statistics/collisions
[  310.456919] Modules linked in: eth1394 floppy ohci1394 ide_cd ieee1394 cdrom
[  310.457259] CPU:    0
[  310.457260] EIP:    0060:[<c0150c9a>]    Not tainted VLI
[  310.457261] EFLAGS: 00210246   (2.6.20-rc1-mm1 #207)
[  310.457478] EIP is at shrink_slab+0x9e/0x169

Looks like we have a problem with slab shrinking here.

Could you please use gdb to check what exactly is at shrink_slab+0x9e?

Quoted 37 lines

[  310.457548] eax: 00000000   ebx: 00000000   ecx: 00000000   edx: 00000000
[  310.457623] esi: 00000000   edi: c18fe500   ebp: f7b3fe3c   esp: f7b3fe08
[  310.457696] ds: 007b   es: 007b   fs: 00d8  gs: 0033  ss: 0068
[  310.457772] Process swsusp (pid: 3243, ti=f7b3e000 task=f756f030
task.ti=f7b3e000)
[  310.457845] Stack: c175d8a0 c175daa0 c175db00 00000000 00000000 c053cec0
000045ec 000000d0
[  310.458286]        00000000 00000000 00001179 00001179 00000000 f7b3fe94
c0151445 00000001
[  310.458723]        f7b3fe64 00001df1 00000002 00000001 00000001 00038000
00000c79 0000117b
[  310.459199] Call Trace:
[  310.459334]  [<c0103f1b>] show_trace_log_lvl+0x1a/0x30
[  310.459450]  [<c0103fd6>] show_stack_log_lvl+0xa5/0xca
[  310.459562]  [<c01041ce>] show_registers+0x1d3/0x2b8
[  310.459673]  [<c01043d4>] die+0x121/0x243
[  310.459781]  [<c010456c>] do_trap+0x76/0x9c
[  310.459892]  [<c0104bd8>] do_divide_error+0x94/0x9e
[  310.460001]  [<c038a7e4>] error_code+0x7c/0x84
[  310.460113]  [<c0151445>] shrink_all_memory+0x211/0x2eb
[  310.460225]  [<c01418c1>] swsusp_shrink_memory+0x187/0x196
[  310.460335]  [<c0141a07>] prepare_processes+0x35/0xc8
[  310.460446]  [<c0141cce>] pm_suspend_disk+0xd/0x16f
[  310.460558]  [<c0140c87>] enter_state+0x129/0x19b
[  310.460668]  [<c0140d9c>] state_store+0xa3/0xac
[  310.460777]  [<c0198ab0>] subsys_attr_store+0x20/0x25
[  310.460889]  [<c0198b9f>] sysfs_write_file+0x97/0xd8
[  310.460998]  [<c0165262>] vfs_write+0x8b/0x149
[  310.461108]  [<c01658cb>] sys_write+0x3d/0x64
[  310.461216]  [<c0102fe4>] syscall_call+0x7/0xb
[  310.461328]  =======================
[  310.461397] Code: 31 c0 ff 17 89 c3 8b 45 e4 31 d2 f7 77 0c f7 e3 89 45 d8 89
55 dc 89 d1 89 c6 31 d2 85 c9 74 09 89 c8 31 d2 f7 75 f0 89 c1 89 f0 <f7> 75 f0
89 ca 89 45 d8 89 55 dc 8b 45 d8 03 47 10 89 47 10 85
[  310.464079] EIP: [<c0150c9a>] shrink_slab+0x9e/0x169 SS:ESP 0068:f7b3fe08
[  310.464228]

Quoted 5 lines

swsusp script is something like this:
echo platform > /sys/power/disk
echo disk > /sys/power/state
regards,

Greetings,
Rafael

Signature · 3 lines

If you don't have the time to read,
you don't have the time or the tools to write.
		- Stephen King

Jiri Slaby

author

2006-12-18 17:03 UTC #2 lore raw

Quoted 20 lines · Rafael J. Wysocki

Hi,

Quoted 16 lines · Jiri Slaby

Hi.

I got this oops while suspending:

[  309.366557] Disabling non-boot CPUs ...
[  309.386563] CPU 1 is now offline
[  309.387625] CPU1 is down
[  309.387704] Stopping tasks ... done.
[  310.030991] Shrinking memory... -<0>divide error: 0000 [#1]
[  310.456669] SMP
[  310.456814] last sysfs file:
/devices/pci0000:00/0000:00:1e.0/0000:02:08.0/eth0/statistics/collisions
[  310.456919] Modules linked in: eth1394 floppy ohci1394 ide_cd ieee1394 cdrom
[  310.457259] CPU:    0
[  310.457260] EIP:    0060:[<c0150c9a>]    Not tainted VLI
[  310.457261] EFLAGS: 00210246   (2.6.20-rc1-mm1 #207)
[  310.457478] EIP is at shrink_slab+0x9e/0x169

Quoted 3 lines

Looks like we have a problem with slab shrinking here.
Could you please use gdb to check what exactly is at shrink_slab+0x9e?

Sure, but not till Friday, sorry (I am away).

Quoted 37 lines

[  310.457548] eax: 00000000   ebx: 00000000   ecx: 00000000   edx: 00000000
[  310.457623] esi: 00000000   edi: c18fe500   ebp: f7b3fe3c   esp: f7b3fe08
[  310.457696] ds: 007b   es: 007b   fs: 00d8  gs: 0033  ss: 0068
[  310.457772] Process swsusp (pid: 3243, ti=f7b3e000 task=f756f030
task.ti=f7b3e000)
[  310.457845] Stack: c175d8a0 c175daa0 c175db00 00000000 00000000 c053cec0
000045ec 000000d0
[  310.458286]        00000000 00000000 00001179 00001179 00000000 f7b3fe94
c0151445 00000001
[  310.458723]        f7b3fe64 00001df1 00000002 00000001 00000001 00038000
00000c79 0000117b
[  310.459199] Call Trace:
[  310.459334]  [<c0103f1b>] show_trace_log_lvl+0x1a/0x30
[  310.459450]  [<c0103fd6>] show_stack_log_lvl+0xa5/0xca
[  310.459562]  [<c01041ce>] show_registers+0x1d3/0x2b8
[  310.459673]  [<c01043d4>] die+0x121/0x243
[  310.459781]  [<c010456c>] do_trap+0x76/0x9c
[  310.459892]  [<c0104bd8>] do_divide_error+0x94/0x9e
[  310.460001]  [<c038a7e4>] error_code+0x7c/0x84
[  310.460113]  [<c0151445>] shrink_all_memory+0x211/0x2eb
[  310.460225]  [<c01418c1>] swsusp_shrink_memory+0x187/0x196
[  310.460335]  [<c0141a07>] prepare_processes+0x35/0xc8
[  310.460446]  [<c0141cce>] pm_suspend_disk+0xd/0x16f
[  310.460558]  [<c0140c87>] enter_state+0x129/0x19b
[  310.460668]  [<c0140d9c>] state_store+0xa3/0xac
[  310.460777]  [<c0198ab0>] subsys_attr_store+0x20/0x25
[  310.460889]  [<c0198b9f>] sysfs_write_file+0x97/0xd8
[  310.460998]  [<c0165262>] vfs_write+0x8b/0x149
[  310.461108]  [<c01658cb>] sys_write+0x3d/0x64
[  310.461216]  [<c0102fe4>] syscall_call+0x7/0xb
[  310.461328]  =======================
[  310.461397] Code: 31 c0 ff 17 89 c3 8b 45 e4 31 d2 f7 77 0c f7 e3 89 45 d8 89
55 dc 89 d1 89 c6 31 d2 85 c9 74 09 89 c8 31 d2 f7 75 f0 89 c1 89 f0 <f7> 75 f0
89 ca 89 45 d8 89 55 dc 8b 45 d8 03 47 10 89 47 10 85
[  310.464079] EIP: [<c0150c9a>] shrink_slab+0x9e/0x169 SS:ESP 0068:f7b3fe08
[  310.464228]

Quoted 3 lines

Quoted 3 lines
swsusp script is something like this:
echo platform > /sys/power/disk
echo disk > /sys/power/state

regards,

Signature · 4 lines

http://www.fi.muni.cz/~xslaby/            Jiri Slaby
faculty of informatics, masaryk university, brno, cz
e-mail: jirislaby gmail com, gpg pubkey fingerprint:
B674 9967 0407 CE62 ACC8  22A0 32CC 55C3 39D4 7A7E

Andrew Morton

2006-12-18 20:59 UTC #3 lore raw

Quoted 21 lines · On Mon, 18 Dec 2006 18:02:20 +0100 Jiri Slaby <[email protected]> wrote:

Quoted 20 lines · Rafael J. Wysocki

Hi,

Quoted 16 lines · Jiri Slaby

Hi.

I got this oops while suspending:

[  309.366557] Disabling non-boot CPUs ...
[  309.386563] CPU 1 is now offline
[  309.387625] CPU1 is down
[  309.387704] Stopping tasks ... done.
[  310.030991] Shrinking memory... -<0>divide error: 0000 [#1]
[  310.456669] SMP
[  310.456814] last sysfs file:
/devices/pci0000:00/0000:00:1e.0/0000:02:08.0/eth0/statistics/collisions
[  310.456919] Modules linked in: eth1394 floppy ohci1394 ide_cd ieee1394 cdrom
[  310.457259] CPU:    0
[  310.457260] EIP:    0060:[<c0150c9a>]    Not tainted VLI
[  310.457261] EFLAGS: 00210246   (2.6.20-rc1-mm1 #207)
[  310.457478] EIP is at shrink_slab+0x9e/0x169

Quoted 5 lines

Quoted 3 lines
Looks like we have a problem with slab shrinking here.
Could you please use gdb to check what exactly is at shrink_slab+0x9e?
Sure, but not till Friday, sorry (I am away).

I think there's only one divide in there which can do this, so...

mm/vmscan.c renamed from mm/vmscan.c~shrink_slab-handle-bad-shrinkers +8 -1

@@ -20,6 +20,7 @@
 #include <linux/pagemap.h>
 #include <linux/init.h>
 #include <linux/highmem.h>
+#include <linux/kallsyms.h>
 #include <linux/vmstat.h>
 #include <linux/file.h>
 #include <linux/writeback.h>

@@ -190,7 +191,13 @@ unsigned long shrink_slab(unsigned long 
 		unsigned long total_scan;
 		unsigned long max_pass = (*shrinker->shrinker)(0, gfp_mask);
 
-		delta = (4 * scanned) / shrinker->seeks;
+		if (!shrinker->seeks) {
+			print_symbol("shrinker %s has zero seeks\n",
+				(unsigned long)shrinker->shrinker);
+			delta = (4 * scanned) / DEFAULT_SEEKS;
+		} else {
+			delta = (4 * scanned) / shrinker->seeks;
+		}
 		delta *= max_pass;
 		do_div(delta, lru_pages + 1);
 		shrinker->nr += delta;

A quick grep shows that all set_shrinker() callers are doing the right thing, so something kooky has happened.

Rafael J. Wysocki

2006-12-18 22:36 UTC #4 lore raw

In reply to Jiri Slaby

Quoted 21 lines · Jiri Slaby

Quoted 20 lines · Rafael J. Wysocki

Hi,

Quoted 16 lines · Jiri Slaby

Hi.

I got this oops while suspending:

[  309.366557] Disabling non-boot CPUs ...
[  309.386563] CPU 1 is now offline
[  309.387625] CPU1 is down
[  309.387704] Stopping tasks ... done.
[  310.030991] Shrinking memory... -<0>divide error: 0000 [#1]
[  310.456669] SMP
[  310.456814] last sysfs file:
/devices/pci0000:00/0000:00:1e.0/0000:02:08.0/eth0/statistics/collisions
[  310.456919] Modules linked in: eth1394 floppy ohci1394 ide_cd ieee1394 cdrom
[  310.457259] CPU:    0
[  310.457260] EIP:    0060:[<c0150c9a>]    Not tainted VLI
[  310.457261] EFLAGS: 00210246   (2.6.20-rc1-mm1 #207)
[  310.457478] EIP is at shrink_slab+0x9e/0x169

Quoted 5 lines

Quoted 3 lines
Looks like we have a problem with slab shrinking here.
Could you please use gdb to check what exactly is at shrink_slab+0x9e?
Sure, but not till Friday, sorry (I am away).

I reproduced this on one box, but then it turned out that EIP was at line 195 of mm/vmscan.c where there was

do_div(delta, lru_pages + 1);

Well, I have no idea how this can lead to a divide error (lru_pages is unsigned).

I'm unable to reproduce this on another i386 box, so it seems to be somewhat configuration specific.

Does 2.6.20-rc1 work for you?

Rafael

Signature · 3 lines

If you don't have the time to read,
you don't have the time or the tools to write.
		- Stephen King

Nigel Cunningham

2006-12-18 22:44 UTC #5 lore raw

Hi.

Quoted 33 lines · Rafael J. Wysocki

Quoted 21 lines · Jiri Slaby
Quoted 20 lines · Rafael J. Wysocki
Hi,
Quoted 16 lines · Jiri Slaby
Hi.
I got this oops while suspending:
[  309.366557] Disabling non-boot CPUs ...
[  309.386563] CPU 1 is now offline
[  309.387625] CPU1 is down
[  309.387704] Stopping tasks ... done.
[  310.030991] Shrinking memory... -<0>divide error: 0000 [#1]
[  310.456669] SMP
[  310.456814] last sysfs file:
/devices/pci0000:00/0000:00:1e.0/0000:02:08.0/eth0/statistics/collisions
[  310.456919] Modules linked in: eth1394 floppy ohci1394 ide_cd ieee1394 cdrom
[  310.457259] CPU:    0
[  310.457260] EIP:    0060:[<c0150c9a>]    Not tainted VLI
[  310.457261] EFLAGS: 00210246   (2.6.20-rc1-mm1 #207)
[  310.457478] EIP is at shrink_slab+0x9e/0x169
Quoted 5 lines
Quoted 3 lines
Looks like we have a problem with slab shrinking here.
Could you please use gdb to check what exactly is at shrink_slab+0x9e?
Sure, but not till Friday, sorry (I am away).
I reproduced this on one box, but then it turned out that EIP was at line 195 of mm/vmscan.c where there was
do_div(delta, lru_pages + 1);

Quoted 7 lines

Well, I have no idea how this can lead to a divide error (lru_pages is unsigned).
I'm unable to reproduce this on another i386 box, so it seems to be somewhat configuration specific.
Does 2.6.20-rc1 work for you?

I have a patch in -mm that reduces lru_pages by what shrink_all_zones returns. Could shrink_all_zones perhaps be returning incorrect values such that lru_pages ends up becoming -1?

Regards,

Nigel

Rafael J. Wysocki

2006-12-18 23:07 UTC #6 lore raw

Hi,

Quoted 39 lines · Nigel Cunningham

Hi.
Quoted 36 lines · Rafael J. Wysocki
Quoted 21 lines · Jiri Slaby
Quoted 20 lines · Rafael J. Wysocki
Hi,
Quoted 16 lines · Jiri Slaby
Hi.
I got this oops while suspending:
[  309.366557] Disabling non-boot CPUs ...
[  309.386563] CPU 1 is now offline
[  309.387625] CPU1 is down
[  309.387704] Stopping tasks ... done.
[  310.030991] Shrinking memory... -<0>divide error: 0000 [#1]
[  310.456669] SMP
[  310.456814] last sysfs file:
/devices/pci0000:00/0000:00:1e.0/0000:02:08.0/eth0/statistics/collisions
[  310.456919] Modules linked in: eth1394 floppy ohci1394 ide_cd ieee1394 cdrom
[  310.457259] CPU:    0
[  310.457260] EIP:    0060:[<c0150c9a>]    Not tainted VLI
[  310.457261] EFLAGS: 00210246   (2.6.20-rc1-mm1 #207)
[  310.457478] EIP is at shrink_slab+0x9e/0x169
Quoted 5 lines
Quoted 3 lines
Looks like we have a problem with slab shrinking here.
Could you please use gdb to check what exactly is at shrink_slab+0x9e?
Sure, but not till Friday, sorry (I am away).
I reproduced this on one box, but then it turned out that EIP was at line 195 of mm/vmscan.c where there was
do_div(delta, lru_pages + 1);
Well, I have no idea how this can lead to a divide error (lru_pages is unsigned).

Quoted 8 lines

Quoted 4 lines
I'm unable to reproduce this on another i386 box, so it seems to be somewhat configuration specific.
Does 2.6.20-rc1 work for you?
I have a patch in -mm that reduces lru_pages by what shrink_all_zones returns. Could shrink_all_zones perhaps be returning incorrect values such that lru_pages ends up becoming -1?

I don't think so, but look at the appended patch. ;-)

Greetings,
Rafael

---
Fix a (really bad) typo in shrink_all_memory().

Signed-off-by Rafael J. Wysocki <[email protected]>

---
 mm/vmscan.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

mm/vmscan.c +1 -1

Index: linux-2.6.20-rc1-mm1/mm/vmscan.c
===================================================================
@@ -1569,7 +1569,7 @@ unsigned long shrink_all_memory(unsigned
 			sc.swap_cluster_max = nr_pages - ret;
 			freed = shrink_all_zones(nr_to_scan, prio, pass, &sc);
 			ret += freed;
-			lru_pages =- freed;
+			lru_pages -= freed;
 			nr_to_scan = nr_pages - ret;
 			if (ret >= nr_pages)
 				goto out;

Andrew Morton

2006-12-18 23:17 UTC #8 lore raw

In reply to Rafael J. Wysocki

Quoted 10 lines · On Mon, 18 Dec 2006 23:38:23 +0100 "Rafael J. Wysocki" <[email protected]> wrote:

Quoted 5 lines
Quoted 3 lines
Looks like we have a problem with slab shrinking here.
Could you please use gdb to check what exactly is at shrink_slab+0x9e?
Sure, but not till Friday, sorry (I am away).
I reproduced this on one box, but then it turned out that EIP was at line 195 of mm/vmscan.c where there was
do_div(delta, lru_pages + 1);

That implies that we passed it lru_pages=-1.

Presumably the logic in vmscanc-account-for-memory-already-freed-in-seeking-to.patch caused that.

Quoted 6 lines

Well, I have no idea how this can lead to a divide error (lru_pages is unsigned).
I'm unable to reproduce this on another i386 box, so it seems to be somewhat configuration specific.

There is one wart in shrink_all_memory() and I think we should fix that in 2.6.20.

Please check the below. I'll drop vmscanc-account-for-memory-already-freed-in-seeking-to.patch. It has other stuff in it which we might still need. But altering sc->swap_cluster_max in that manner looks odd.

From: Andrew Morton <[email protected]>

At the end of shrink_all_memory() we forget to recalculate lru_pages: it can be zero.

Fix that up, and add a helper function for this operation too.

Also, recalculate lru_pages each time around the inner loop to get the balancing correct.

Cc "Rafael J. Wysocki" <[email protected]>

Signed-off-by Andrew Morton <[email protected]>

---

 mm/vmscan.c |   33 ++++++++++++++++-----------------
 1 files changed, 16 insertions(+), 17 deletions(-)

mm/vmscan.c renamed from mm/vmscan.c~shrink_all_memory-fix-lru_pages-handling +16 ~~-17~~

diff -puN mm/vmscan.c~shrink_all_memory-fix-lru_pages-handling mm/vmscan.c
@@ -1484,6 +1484,16 @@ static unsigned long shrink_all_zones(un
 	return ret;
 }
 
+static unsigned long count_lru_pages(void)
+{
+	struct zone *zone;
+	unsigned long ret = 0;
+
+	for_each_zone(zone);
+		ret += zone->nr_active + zone->nr_inactive;
+	return ret;
+}
+
 /*
  * Try to free `nr_pages' of memory, system-wide, and return the number of
  * freed pages.

@@ -1498,7 +1508,6 @@ unsigned long shrink_all_memory(unsigned
 	unsigned long ret = 0;
 	int pass;
 	struct reclaim_state reclaim_state;
-	struct zone *zone;
 	struct scan_control sc = {
 		.gfp_mask = GFP_KERNEL,
 		.may_swap = 0,

@@ -1509,10 +1518,7 @@ unsigned long shrink_all_memory(unsigned
 
 	current->reclaim_state = &reclaim_state;
 
-	lru_pages = 0;
-	for_each_zone(zone)
-		lru_pages += zone->nr_active + zone->nr_inactive;
-
+	lru_pages = count_lru_pages();
 	nr_slab = global_page_state(NR_SLAB_RECLAIMABLE);
 	/* If slab caches are huge, it's better to hit them first */
 	while (nr_slab >= lru_pages) {

@@ -1539,13 +1545,6 @@ unsigned long shrink_all_memory(unsigned
 	for (pass = 0; pass < 5; pass++) {
 		int prio;
 
-		/* Needed for shrinking slab caches later on */
-		if (!lru_pages)
-			for_each_zone(zone) {
-				lru_pages += zone->nr_active;
-				lru_pages += zone->nr_inactive;
-			}
-
 		/* Force reclaiming mapped pages in the passes #3 and #4 */
 		if (pass > 2) {
 			sc.may_swap = 1;

@@ -1561,7 +1560,8 @@ unsigned long shrink_all_memory(unsigned
 				goto out;
 
 			reclaim_state.reclaimed_slab = 0;
-			shrink_slab(sc.nr_scanned, sc.gfp_mask, lru_pages);
+			shrink_slab(sc.nr_scanned, sc.gfp_mask,
+					count_lru_pages());
 			ret += reclaim_state.reclaimed_slab;
 			if (ret >= nr_pages)
 				goto out;

@@ -1569,20 +1569,19 @@ unsigned long shrink_all_memory(unsigned
 			if (sc.nr_scanned && prio < DEF_PRIORITY - 2)
 				congestion_wait(WRITE, HZ / 10);
 		}
-
-		lru_pages = 0;
 	}
 
 	/*
 	 * If ret = 0, we could not shrink LRUs, but there may be something
 	 * in slab caches
 	 */
-	if (!ret)
+	if (!ret) {
 		do {
 			reclaim_state.reclaimed_slab = 0;
-			shrink_slab(nr_pages, sc.gfp_mask, lru_pages);
+			shrink_slab(nr_pages, sc.gfp_mask, count_lru_pages());
 			ret += reclaim_state.reclaimed_slab;
 		} while (ret < nr_pages && reclaim_state.reclaimed_slab > 0);
+	}
 
 out:
 	current->reclaim_state = NULL;

Rafael J. Wysocki

2006-12-19 00:50 UTC #9 lore raw

Quoted 22 lines · Andrew Morton

Quoted 10 lines · On Mon, 18 Dec 2006 23:38:23 +0100 "Rafael J. Wysocki" <[email protected]> wrote:
Quoted 5 lines
Quoted 3 lines
Looks like we have a problem with slab shrinking here.
Could you please use gdb to check what exactly is at shrink_slab+0x9e?
Sure, but not till Friday, sorry (I am away).
I reproduced this on one box, but then it turned out that EIP was at line 195 of mm/vmscan.c where there was
do_div(delta, lru_pages + 1);
That implies that we passed it lru_pages=-1.
Presumably the logic in vmscanc-account-for-memory-already-freed-in-seeking-to.patch caused that.
Quoted 3 lines
Well, I have no idea how this can lead to a divide error (lru_pages is unsigned).

Quoted 8 lines

Quoted 3 lines
I'm unable to reproduce this on another i386 box, so it seems to be somewhat configuration specific.
There is one wart in shrink_all_memory() and I think we should fix that in 2.6.20.
Please check the below.

Fine by me.

Quoted 3 lines

I'll drop vmscanc-account-for-memory-already-freed-in-seeking-to.patch. It has other stuff in it which we might still need. But altering sc->swap_cluster_max in that manner looks odd.

Agreed.

Greetings,
Rafael

David Rientjes

2006-12-19 01:18 UTC #10 lore raw

In reply to Andrew Morton

Quoted 10 lines · Andrew Morton

diff -puN mm/vmscan.c~shrink_all_memory-fix-lru_pages-handling mm/vmscan.c
mm/vmscan.c renamed from mm/vmscan.c~shrink_all_memory-fix-lru_pages-handling +3 -0
@@ -1484,6 +1484,16 @@ static unsigned long shrink_all_zones(un
 	return ret;
 }
 
+static unsigned long count_lru_pages(void)
+{
+	struct zone *zone;

mm/vmscan.c static unsigned long shrink_all_zones(un · line 1499

Quoted 10 lines

mm/vmscan.c renamed from mm/vmscan.c~shrink_all_memory-fix-lru_pages-handling +7 -0

+	unsigned long ret = 0;
+
+	for_each_zone(zone);
+		ret += zone->nr_active + zone->nr_inactive;
+	return ret;
+}
+
 /*
  * Try to free `nr_pages' of memory, system-wide, and return the number of
  * freed pages.

There's an extra semicolon there that results in only the final zone being used.

David

Andrew Morton

2006-12-19 01:28 UTC #11 lore raw

Quoted 9 lines · On Mon, 18 Dec 2006 17:18:12 -0800 (PST) David Rientjes…

Quoted 7 lines · Andrew Morton
diff -puN mm/vmscan.c~shrink_all_memory-fix-lru_pages-handling mm/vmscan.c
mm/vmscan.c renamed from mm/vmscan.c~shrink_all_memory-fix-lru_pages-handling +0 -0
@@ -1484,6 +1484,16 @@ static unsigned long shrink_all_zones(un
 	return ret;
 }
 

Quoted 15 lines

mm/vmscan.c static unsigned long shrink_all_zones(un · line 1499
Quoted 13 lines
mm/vmscan.c renamed from mm/vmscan.c~shrink_all_memory-fix-lru_pages-handling +10 -0
+static unsigned long count_lru_pages(void)
+{
+	struct zone *zone;
+	unsigned long ret = 0;
+
+	for_each_zone(zone);
+		ret += zone->nr_active + zone->nr_inactive;
+	return ret;
+}
+
 /*
  * Try to free `nr_pages' of memory, system-wide, and return the number of
  * freed pages.
There's an extra semicolon there

Sigh. coding-while-diseased.

Quoted 3 lines

that results in only the final zone being
used.

Actually it'll go oops. Fixed, thanks.