Use osreldate to check the existence of getrandom(2).
ClosedPublic
Actions

Authored by delphij on Aug 20 2018, 12:44 AM.

Details

Reviewers

cem
kib

Commits

rS338084: In r331279 the code used ENOSYS to check the existence of getrandom(2).

Summary

In r331279 the code used ENOSYS to check the existence of getrandom(2).
This will only work if the caller already handles SIGSYS, which is not
always the case.

Address this by checking osreldate instead. Note that because there
was not __FreeBSD_version bump when the system call was added, use
1200061 (r332100) which is the first bump after the introduction of
the system call.

PR: 230762

Test Plan

Run the code against a kernel older than r331279.

Diff Detail

Repository

rS FreeBSD src repository - subversion

Lint

Lint Not Applicable

Unit

Tests Not Applicable

Event Timeline

delphij created this revision.Aug 20 2018, 12:44 AM

Herald added a subscriber: imp. · View Herald TranscriptAug 20 2018, 12:44 AM

Harbormaster completed remote builds in B18958: Diff 46949.Aug 20 2018, 12:44 AM

Good find, thank you. Fix looks good to me.

lib/libc/gen/getentropy.c
113 ↗	(On Diff #46949)	We have lost our ENOSYS handling for programs that mask SIGSYS and are in the small range of revisions between introduction of the syscall and GETRANDOM_FIRST, but I think that is probably ok.
135–136 ↗	(On Diff #46949)	I'd suggest reordering this as `if (!have_getrandom) return (...fallback());` to avoid the extra indentation in the normal case.

This revision is now accepted and ready to land.Aug 20 2018, 1:03 AM

delphij marked an inline comment as done.Aug 20 2018, 1:56 AM

delphij added inline comments.

lib/libc/gen/getentropy.c
113 ↗	(On Diff #46949)	Yeah, the only way to do it would be to create a SIGSYS handler for >= 1200060 and < 1200061, but since it's -CURRENT it's probably not worth the effort.
135–136 ↗	(On Diff #46949)	I think it's against style(9). Let's revisit this at a later time.