Skip to contentSkip to conversation

❌ FAIL: Waiting for review: Test report for kernel 5.6.14-dbf782f.cki (stable-queue)

LTP Li Wang 2 participants 2 messages May 26, 2020 4 messages not in the archive Library code

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): CAEemH2dNB3Ktt5LeBLxMKekv7-gj4d=EkQouYmO1SgbyzjLEaw@mail.gmail.com

A message that is not in the archive (sent off-list or not yet received): [email protected]

Li Wang

author
#4 lore raw
Quoted 25 lines · Jan Stancek

...

Quoted 8 lines
Quoted 7 lines
Quoted 2 lines
    s390x:
     ? LTP

@Li, any thoughts on hugetlb TBROKs here? I thought you said 'request_hugepages'
was not supposed to be hard test requirement.

Quoted 6 lines

The hugemmap05.c doesn't goes the new 'request_hugepages' way in the huge page reserving. it tries to reserve 128 hpages and overcommits 192 hpages in hard coding, and probably touch the memory available uplimit on s390x. static long size = 128, length = 384;
...
addr = SAFE_MMAP(ADDR, (length / 2 * hugepagesize), PROTECTION, FLAGS,

fd,

Quoted 5 lines

0);

We could convert this to use the new request_hugepages API and do the requested number of huge page checking at setup() to make sure the system only running in a reasonable situation.

Quoted 6 lines

Sorry, I didn't make it clear I was talking about hugemmap06. The patch is OK,
since it makes both use request_hugepages API. But should we also do something
with this TBROK?

It seems unkind and useless to exit with TBROK when failed to reserve with 80% max_hpages again. I'm thinking if it could be directly replaced by TCONF to skip the whole test?

lib/tst_hugepage.c +0 -1
@@ -43,7 +43,7 @@ unsigned long tst_request_hugepages(unsigned long hpages)
        SAFE_FILE_PRINTF(PATH_NR_HPAGES, "%lu", tst_hugepages);
        SAFE_FILE_SCANF(PATH_NR_HPAGES, "%lu", &val);
        if (val != tst_hugepages)
-               tst_brk(TBROK, "nr_hugepages = %lu, but expect %lu", val,

tst_hugepages);
+ tst_brk(TCONF, "Not enough hugepages for testing.");

        tst_res(TINFO, "%lu hugepage(s) reserved", tst_hugepages);

out:

Quoted 4 lines
    tst_hugepage.c:46: BROK: nr_hugepages = 171, but expect 255
Signature · 5 lines
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20200526/1ba02bf6/attachment.htm>
#5 lore raw

----- Original Message -----

Quoted 34 lines · Jan Stancek
Quoted 26 lines

...

Quoted 8 lines
Quoted 7 lines
Quoted 2 lines
    s390x:
     ? LTP

@Li, any thoughts on hugetlb TBROKs here? I thought you said 'request_hugepages'
was not supposed to be hard test requirement.

Quoted 6 lines

The hugemmap05.c doesn't goes the new 'request_hugepages' way in the huge page reserving. it tries to reserve 128 hpages and overcommits 192 hpages in hard coding, and probably touch the memory available uplimit on s390x. static long size = 128, length = 384;
...
addr = SAFE_MMAP(ADDR, (length / 2 * hugepagesize), PROTECTION, FLAGS,

fd,

Quoted 5 lines

0);

We could convert this to use the new request_hugepages API and do the requested number of huge page checking at setup() to make sure the system only running in a reasonable situation.

Quoted 6 lines

Sorry, I didn't make it clear I was talking about hugemmap06. The patch is OK,
since it makes both use request_hugepages API. But should we also do something
with this TBROK?

Quoted 3 lines

It seems unkind and useless to exit with TBROK when failed to reserve with 80% max_hpages again. I'm thinking if it could be directly replaced by TCONF to skip the whole test?

I think it can fail even when request is less than 80%, when memory is too fragmented. I agree with TCONF, but I'd keep numbers (val, tst_hugepages) in output, so we know how close it was to the requested number.

Quoted 24 lines
lib/tst_hugepage.c +1 -1
@@ -43,7 +43,7 @@ unsigned long tst_request_hugepages(unsigned long hpages)
        SAFE_FILE_PRINTF(PATH_NR_HPAGES, "%lu", tst_hugepages);
        SAFE_FILE_SCANF(PATH_NR_HPAGES, "%lu", &val);
        if (val != tst_hugepages)
-               tst_brk(TBROK, "nr_hugepages = %lu, but expect %lu", val,
 tst_hugepages);
+               tst_brk(TCONF, "Not enough hugepages for testing.");
 
        tst_res(TINFO, "%lu hugepage(s) reserved", tst_hugepages);

out:

Quoted 4 lines
    tst_hugepage.c:46: BROK: nr_hugepages = 171, but expect 255

--
Regards,
Li Wang