sendfile09 is already slow on my machine with SSD (2-5 seconds). So I can imagine how it exceeds the 30-second default timeout on our tiny NUCs with probably super-slow HDDs. Also recall that LTP tests are run in parallel in our CI, so the perf gets even worse.
I checked other sendfile* LTP tests, and they all operate on small data (up to 100 bytes). That's why they never timeout, but sendfile09 does.
UPDATE: Added fsync02 here, with timeout of 120 seconds. Looking at the sources of this test, I see this:
/* fsync() has to finish within TIME_LIMIT. */
#define TIME_LIMIT 120
How to test this PR?
Jenkins. Can try several times to make sure we solved this issue.
Description of the changes
sendfile09
is already slow on my machine with SSD (2-5 seconds). So I can imagine how it exceeds the 30-second default timeout on our tiny NUCs with probably super-slow HDDs. Also recall that LTP tests are run in parallel in our CI, so the perf gets even worse.I checked other
sendfile*
LTP tests, and they all operate on small data (up to 100 bytes). That's why they never timeout, butsendfile09
does.UPDATE: Added
fsync02
here, with timeout of 120 seconds. Looking at the sources of this test, I see this:How to test this PR?
Jenkins. Can try several times to make sure we solved this issue.
This change is