Page MenuHomeFreeBSD

nvmecontrol: Refactor devlist implementation
ClosedPublic

Authored by jhb on Oct 31 2024, 2:52 PM.
Tags
None
Referenced Files
F147509318: D47354.id146020.diff
Wed, Mar 11, 12:58 PM
F147486555: D47354.id145735.diff
Wed, Mar 11, 9:11 AM
F147469982: D47354.id145735.diff
Wed, Mar 11, 6:32 AM
F147469947: D47354.id145735.diff
Wed, Mar 11, 6:31 AM
Unknown Object (File)
Tue, Mar 10, 11:42 AM
Unknown Object (File)
Tue, Mar 10, 8:00 AM
Unknown Object (File)
Tue, Mar 10, 7:05 AM
Unknown Object (File)
Tue, Mar 10, 12:07 AM
Subscribers

Details

Summary

Split out helper functions for scan_namespace and scan_controller.
While here, replace sprintf() calls with snprintf() and avoid
leaking the contoller fd if read_controller_data() fails.

Sponsored by: Chelsio Communications

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Not Applicable
Unit
Tests Not Applicable

Event Timeline

jhb requested review of this revision.Oct 31 2024, 2:52 PM

This makes things a lot better. But I can't help but think scanning sequential unit numbers couldn't be better maybe.

sbin/nvmecontrol/devlist.c
95

Is there a better #define that we can use? 64 is fine, but I can't say why

This revision is now accepted and ready to land.Oct 31 2024, 10:14 PM
sbin/nvmecontrol/devlist.c
95

That is in the old code and I just kept it. 64 is arbitrary, but as long as it's only in one place and everything else uses sizeof(name) I'm fine with that. If we had a constant for the largest possible filename in a component, maybe that could be used instead? (So not PATH_MAX, but NAME_MAX I guess?)

sbin/nvmecontrol/devlist.c
95

Gotcha!

This revision was automatically updated to reflect the committed changes.