[LTP] [PATCH] swapon01: swapon01: prevent OOM happening in swap process
Wei Gao
wegao@suse.com
Mon Mar 18 13:39:47 CET 2024
On Mon, Mar 18, 2024 at 03:26:00PM +0800, Li Wang wrote:
> On Mon, Mar 18, 2024 at 3:02 PM Wei Gao <wegao@suse.com> wrote:
>
> > On Mon, Mar 18, 2024 at 02:32:41PM +0800, Li Wang wrote:
> > > On Mon, Mar 18, 2024 at 11:20 AM Wei Gao <wegao@suse.com> wrote:
> > >
> > > > On Sun, Mar 17, 2024 at 05:52:01PM +0800, Li Wang wrote:
> > > > > Hi Wei,
> > > > >
> > > > > Can you try this one and post the test log here?
> > > > >
> > > > > And again, it'd be helpful to know the config of your SUT.
> > > > > e.g. `free -h` `lscpu` `uname -r` infoformation
> > > > >
> > > > >
> > > > > --- a/testcases/kernel/syscalls/swapon/swapon01.c
> > > > > +++ b/testcases/kernel/syscalls/swapon/swapon01.c
> > > > > @@ -37,11 +37,20 @@ static void verify_swapon(void)
> > > > >
> > > > > static void setup(void)
> > > > > {
> > > > > + tst_enable_oom_protection(0);
> > > > > is_swap_supported(SWAP_FILE);
> > > > > - make_swapfile(SWAP_FILE, 10, 0);
> > > > > + make_swapfile(SWAP_FILE, 1024, 0);
> > > > >
> > > > > SAFE_CG_PRINTF(tst_cg, "cgroup.procs", "%d", getpid());
> > > > > SAFE_CG_PRINTF(tst_cg, "memory.max", "%lu", TESTMEM);
> > > > > +
> > > > > + if (SAFE_CG_HAS(tst_cg, "memory.swap.max"))
> > > > > + SAFE_CG_PRINTF(tst_cg, "memory.swap.max", "%ld",
> > TESTMEM
> > > > *
> > > > > 2);
> > > > > +}
> > > > > +
> > > > > +static void cleanup(void)
> > > > > +{
> > > > > + tst_disable_oom_protection(0);
> > > > > }
> > > > >
> > > > > static struct tst_test test = {
> > > > > @@ -51,5 +60,6 @@ static struct tst_test test = {
> > > > > .all_filesystems = 1,
> > > > > .needs_cgroup_ctrls = (const char *const []){ "memory", NULL
> > },
> > > > > .test_all = verify_swapon,
> > > > > - .setup = setup
> > > > > + .setup = setup,
> > > > > + .cleanup = cleanup
> > > > > };
> > > > >
> > > > >
> > >
> > >
> > >
> > > >
> > > > localhost:~ # free -h <<<< before running case
> > > > total used free shared buff/cache
> > > > available
> > > > Mem: 3.8Gi 478Mi 3.3Gi 9.0Mi 346Mi
> > > > 3.4Gi
> > > > Swap: 0B 0B 0B
> > > >
> > >
> > > I see, it is very likely that your system has no swap space
> > > so that the TESTMEM(1GB) can not be swapped out, then
> > > OOM killer has to be called out.
> > >
> > > I can reproduce your problem by disabling all swap files on my system.
> > >
> > > So the possible fix way is to reduce the TESTMEM size and increase
> > > the tested swapfile.
> > >
> > > I guess this patch can work for you, have a try?
> >
> > No oom, but SwapCached size is unstable, most of time is still 0
> >
>
>
> That's because the available swapfile on your SUT is too small,
> you can adjust it (then retest it) by yourself to find a proper size.
>
> This is fine as long as the swapfile size is less than 300MB,
> otherwise we need to set .dev_min_size like what we did
> for swapoff01.c.
>
> And, on the other side, we can't guarantee the system SwapCached
> happened every time, it depends on the system's configuration.
>
100M is good enough for current system, could you help check following patch?
--- a/testcases/kernel/syscalls/swapon/swapon01.c
+++ b/testcases/kernel/syscalls/swapon/swapon01.c
@@ -14,13 +14,15 @@
#include <unistd.h>
#include <errno.h>
#include <stdlib.h>
+#include <sys/statvfs.h>
#include "tst_test.h"
#include "lapi/syscalls.h"
#include "libswap.h"
#define MNTPOINT "mntpoint"
#define SWAP_FILE MNTPOINT"/swapfile01"
#define TESTMEM (1UL<<30)
+#define SWAP_SIZE 100 * 1024 * 1024 /* 100MB */
static void verify_swapon(void)
{
@@ -37,8 +39,16 @@ static void verify_swapon(void)
static void setup(void)
{
+ struct statvfs fs_info;
+ unsigned long blk_size;
+
+ if (statvfs(".", &fs_info) == -1)
+ tst_brk(TBROK | TERRNO, "Failed to call statvfs");
+
+ blk_size = fs_info.f_bsize;
+
is_swap_supported(SWAP_FILE);
- make_swapfile(SWAP_FILE, 10, 0);
+ make_swapfile(SWAP_FILE, SWAP_SIZE / blk_size , 0);
SAFE_CG_PRINTF(tst_cg, "cgroup.procs", "%d", getpid());
SAFE_CG_PRINTF(tst_cg, "memory.max", "%lu", TESTMEM);
>
> --
> Regards,
> Li Wang
More information about the ltp
mailing list