| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <1e9383d8-98a7-4ef4-9c30-4f068fc38db1@gmail.com>
Date: Thu, 2 May 2024 20:13:51 +0100
From: Usama Arif <usamaarif642@...il.com>
To: Yosry Ahmed <yosryahmed@...gle.com>
Cc: akpm@...ux-foundation.org, hannes@...xchg.org, nphamcs@...il.com,
chengming.zhou@...ux.dev, linux-mm@...ck.org, linux-kernel@...r.kernel.org,
kernel-team@...a.com
Subject: Re: [PATCH] selftests: cgroup: add tests to verify the zswap
writeback path
On 01/05/2024 18:15, Yosry Ahmed wrote:
> Hi Usama,
>
> On Wed, May 1, 2024 at 3:04 AM Usama Arif <usamaarif642@...il.com> wrote:
>> The condition for writeback can be triggered by allocating random
>> memory more than memory.high to push memory into zswap, more than
>> zswap.max to trigger writeback if enabled, but less than memory.max
>> so that OOM is not triggered. Both values of memory.zswap.writeback
>> are tested.
> Thanks for working on this :)
>
>> Signed-off-by: Usama Arif <usamaarif642@...il.com>
>> ---
>> tools/testing/selftests/cgroup/test_zswap.c | 83 +++++++++++++++++++++
>> 1 file changed, 83 insertions(+)
>>
>> diff --git a/tools/testing/selftests/cgroup/test_zswap.c b/tools/testing/selftests/cgroup/test_zswap.c
>> index f0e488ed90d8..fe0e7221525c 100644
>> --- a/tools/testing/selftests/cgroup/test_zswap.c
>> +++ b/tools/testing/selftests/cgroup/test_zswap.c
>> @@ -94,6 +94,19 @@ static int allocate_bytes(const char *cgroup, void *arg)
>> return 0;
>> }
>>
>> +static int allocate_random_bytes(const char *cgroup, void *arg)
>> +{
>> + size_t size = (size_t)arg;
>> + char *mem = (char *)malloc(size);
>> +
>> + if (!mem)
>> + return -1;
>> + for (int i = 0; i < size; i++)
>> + mem[i] = rand() % 128;
>> + free(mem);
>> + return 0;
>> +}
>> +
>> static char *setup_test_group_1M(const char *root, const char *name)
>> {
>> char *group_name = cg_name(root, name);
>> @@ -248,6 +261,74 @@ static int test_zswapin(const char *root)
>> return ret;
>> }
>>
>> +/* Test to verify the zswap writeback path */
>> +static int test_zswap_writeback(const char *root, bool wb)
>> +{
>> + int ret = KSFT_FAIL;
>> + char *test_group;
>> + long zswpwb_before, zswpwb_after;
>> +
>> + test_group = cg_name(root,
>> + wb ? "zswap_writeback_enabled_test" : "zswap_writeback_disabled_test");
>> + if (!test_group)
>> + goto out;
>> + if (cg_create(test_group))
>> + goto out;
>> + if (cg_write(test_group, "memory.max", "8M"))
>> + goto out;
>> + if (cg_write(test_group, "memory.high", "2M"))
>> + goto out;
>> + if (cg_write(test_group, "memory.zswap.max", "2M"))
>> + goto out;
>> + if (cg_write(test_group, "memory.zswap.writeback", wb ? "1" : "0"))
>> + goto out;
>> +
>> + zswpwb_before = cg_read_key_long(test_group, "memory.stat", "zswpwb ");
>> + if (zswpwb_before < 0) {
>> + ksft_print_msg("failed to get zswpwb_before\n");
>> + goto out;
>> + }
>> +
>> + /*
>> + * Allocate more than memory.high to push memory into zswap,
>> + * more than zswap.max to trigger writeback if enabled,
>> + * but less than memory.max so that OOM is not triggered
>> + */
>> + if (cg_run(test_group, allocate_random_bytes, (void *)MB(3)))
>> + goto out;
> We set the zswap limit to 2M. So for this to work properly we need to
> guarantee that the 3M of random data will compress into more than 2M.
> Is this true for all possible zpool implementations and compression
> algorithms? How likely for this to break and start producing false
> negatives if zswap magically becomes more efficient?
>
> One alternative approach that I used before, although more complex, is
> to start by compressing the memory (i.e. through reclaim) without a
> zswap limit, and check the zswap usage. Then, fault the memory back
> in, set the zswap limit lower than the observed usage, and repeat.
> This should guarantee writeback AFAICT.
>
> Also, using memory.reclaim may be easier than memory.high if you
> follow this approach, as you would need to raise memory.high again to
> be able to decompress the memory.
>
Thanks for the review! I have sent a v2 with the method you described. I
did like the simplicity of the method in this v1 a lot more, and we
could have increased the random memory, which eventhough theoretically
would mean it might not trigger a writeback if there was some new magic
compression method, in practice it would always trigger it and work.
Your suggestion which is in v2 covers both theory and practice :)
>> +
>> + /* Verify that zswap writeback occurred only if writeback was enabled */
>> + zswpwb_after = cg_read_key_long(test_group, "memory.stat", "zswpwb ");
>> + if (wb) {
>> + if (zswpwb_after <= zswpwb_before) {
>> + ksft_print_msg("writeback enabled and zswpwb_after <= zswpwb_before\n");
>> + goto out;
>> + }
>> + } else {
>> + if (zswpwb_after != zswpwb_before) {
>> + ksft_print_msg("writeback disabled and zswpwb_after != zswpwb_before\n");
>> + goto out;
>> + }
>> + }
>> +
>> + ret = KSFT_PASS;
>> +
>> +out:
>> + cg_destroy(test_group);
>> + free(test_group);
>> + return ret;
>> +}
>> +
>> +static int test_zswap_writeback_enabled(const char *root)
>> +{
>> + return test_zswap_writeback(root, true);
>> +}
>> +
>> +static int test_zswap_writeback_disabled(const char *root)
>> +{
>> + return test_zswap_writeback(root, false);
>> +}
>> +
>> /*
>> * When trying to store a memcg page in zswap, if the memcg hits its memory
>> * limit in zswap, writeback should affect only the zswapped pages of that
>> @@ -425,6 +506,8 @@ struct zswap_test {
>> T(test_zswap_usage),
>> T(test_swapin_nozswap),
>> T(test_zswapin),
>> + T(test_zswap_writeback_enabled),
>> + T(test_zswap_writeback_disabled),
>> T(test_no_kmem_bypass),
>> T(test_no_invasive_cgroup_shrink),
>> };
>> --
>> 2.43.0
>>
Powered by blists - more mailing lists