Page MenuHomeFreeBSD

New I/O scheduler for FreeBSD
ClosedPublic

Authored by imp on Dec 18 2015, 4:43 AM.
Tags
None
Referenced Files
Unknown Object (File)
Sat, Apr 20, 11:12 AM
Unknown Object (File)
Wed, Apr 17, 2:20 AM
Unknown Object (File)
Thu, Apr 11, 3:19 PM
Unknown Object (File)
Thu, Apr 11, 7:04 AM
Unknown Object (File)
Mar 21 2024, 2:02 AM
Unknown Object (File)
Mar 21 2024, 1:59 AM
Unknown Object (File)
Mar 15 2024, 11:05 AM
Unknown Object (File)
Mar 11 2024, 10:04 AM

Details

Summary

Plus a lot of other things.

Diff Detail

Repository
rS FreeBSD src repository - subversion
Lint
Lint Passed
Unit
No Test Coverage
Build Status
Buildable 1718
Build 1724: arc lint + arc unit

Event Timeline

imp retitled this revision from to New I/O scheduler for FreeBSD.
imp updated this object.
imp edited the test plan for this revision. (Show Details)
This revision was automatically updated to reflect the committed changes.
ngie added inline comments.
head/sys/cam/ata/ata_da.c
923 ↗(On Diff #15202)
  • Why 5?
  • Why get rid of the parameter?
1300 ↗(On Diff #15202)

Why 16?

head/sys/cam/cam_ccb.h
584 ↗(On Diff #15202)

Why isn't this list sorted?

head/sys/cam/scsi/scsi_da.c
2817 ↗(On Diff #15202)

XXX comment?

2906 ↗(On Diff #15202)

XXX comment in C++?

head/sys/conf/options
332 ↗(On Diff #15202)

Why is this in FreeBSD (a specific NETFLIX constant?)

head/sys/dev/ahci/ahci.c
2420 ↗(On Diff #15202)

Could this be done better (even if it breaks KBI), and if so, is there a plan to do this?

head/sys/cam/ata/ata_da.c
923 ↗(On Diff #15202)

That's a regression. It used to be 5 seconds :)

1300 ↗(On Diff #15202)

Because it's long enough and doesn't warrant its own #define.

head/sys/cam/cam_ccb.h
584 ↗(On Diff #15202)

Because ABI breakage.

head/sys/conf/options
332 ↗(On Diff #15202)

Because it is the Netflix scheduler, as opposed to the default scheduler. That's its name, not a Netflix, Inc specific define.

head/sys/dev/ahci/ahci.c
2420 ↗(On Diff #15202)

Can't be done w/o breaking the KBI between CAM and userland. Ken Merry, Scott Long and I have been working on a more general solution because the Auxiliary register is used in a lot of the SMR commands, and it needs to have values other than 1 or 0. Since I did most of this work in 10.x, I didn't want to break the KBI. So we have this ugly kludge until the more general work is completed.