-
Notifications
You must be signed in to change notification settings - Fork 302
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
DAOS-15120 crt: Misc d_rand-related fixes #13738
Conversation
- Remove duplicate d_srand calls in crt. - Use tv_nsec instead of tv_nsec / 1e3 when calling d_srand, for this reduces the chance of seed collisions among daos_engine processes. - Replace rand calls in libgurt and libpool with d_rand. - Remove d_randn (introduced by this author), since lrand48_r doesn't return the lowest bits from Xi, and the % method is therefore sufficient for our purposes. Signed-off-by: Li Wei <[email protected]> Required-githooks: true
Bug-tracker data: |
d_srand(d_timeus_secdiff(0) + getpid()); | ||
rc = d_gettime(&now); | ||
D_ASSERTF(rc == 0, "d_gettime: " DF_RC "\n", DP_RC(rc)); | ||
d_srand(now.tv_sec * 1000 * 1000 * 1000 + now.tv_nsec + getpid()); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
minor: i feel now.tv_nsec + getpid() is sufficient for seed. wont now.tv_sec * 1000 * 1000 * 1000 cause overflows? coverity might complain if so:)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I didn't make this seeding change until I noticed a couple of identical seeds (printed by a temporary fprintf
that I've deleted from this PR) being used during a few dmg system start
commands of 6 engines. The nanosecond-based seeding seemed to be absent of such collisions. (Actually I would prefer calling getrandom
to obtain truly random bytes, but let's see how the current approach works first.)
I think overflows in this case is okay, for d_srand
uses the lower 32 bits only. Didn't think of Coverity though; if Coverity complains, that's my bad and feel free to assign such issue to me. :)
Pick from all possible ranks for protocol query handling Backports the following patches, last one fixes the issue but prior two are benign and make the backport work without conflict DAOS-13717 client: use d_rand() to generate random number (#12418) DAOS-15120 crt: Misc d_rand-related fixes (#13738) DAOS-15476 client: do not use rsvc for proto query rank selection (#14131) Signed-off-by: Mohamad Chaarawi <[email protected]>
Pick from all possible ranks for protocol query handling Backports the following patches, last one fixes the issue but prior two are benign and make the backport work without conflict DAOS-13717 client: use d_rand() to generate random number (#12418) DAOS-15120 crt: Misc d_rand-related fixes (#13738) DAOS-15476 client: do not use rsvc for proto query rank selection (#14131) Required-githooks: true Change-Id: Icd83ae9307a2dfc928ca524af2954131b161157d Signed-off-by: Mohamad Chaarawi <[email protected]> Signed-off-by: Wang Shilong <[email protected]> Signed-off-by: Li Wei <[email protected]> Signed-off-by: Jeff Olivier <[email protected]>
Pick from all possible ranks for protocol query handling Backports the following patches, last one fixes the issue but prior two are benign and make the backport work without conflict DAOS-13717 client: use d_rand() to generate random number (#12418) DAOS-15120 crt: Misc d_rand-related fixes (#13738) DAOS-15476 client: do not use rsvc for proto query rank selection (#14131) DAOS-15540 cart: Change proto query default timeout (#14382) Signed-off-by: Mohamad Chaarawi <[email protected]> Signed-off-by: Wang Shilong <[email protected]> Signed-off-by: Li Wei <[email protected]> Signed-off-by: Jeff Olivier <[email protected]>
Remove duplicate d_srand calls in crt.
Use tv_nsec instead of tv_nsec / 1e3 when calling d_srand, for this reduces the chance of seed collisions among daos_engine processes.
Replace rand calls in libgurt and libpool with d_rand.
Remove d_randn (introduced by this author), since lrand48_r doesn't return the lowest bits from Xi, and the % method is therefore sufficient for our purposes.
Before requesting gatekeeper:
Features:
(orTest-tag*
) commit pragma was used or there is a reason documented that there are no appropriate tags for this PR.Gatekeeper: