There are some cases of OSD use where the value is only initialized once
at a point where successive access of the value can be done so safely
without the need to take the lock.
Obtained from: Juniper Networks, Inc.
Details
Details
Diff Detail
Diff Detail
- Repository
- rG FreeBSD src repository
- Lint
Lint Skipped - Unit
Tests Skipped - Build Status
Buildable 57717 Build 54605: arc lint + arc unit
Event Timeline
Comment Actions
Some uses of OSD may do so in a way where the value is only set once at a point where successive get of the value can be done so safely without the need to take the lock.
I think this sentence is not quite right. :)
| sys/kern/kern_osd.c | ||
|---|---|---|
| 281 | osd_get_unlocked() would be more consistent with this kind of pattern elsewhere in the tree. | |
| 291 | ||
| sys/kern/kern_osd.c | ||
|---|---|---|
| 291 | Assuming you just didn't see it, I think this suggestion should still be applied. Or is there some reason not to? | |
| sys/kern/kern_osd.c | ||
|---|---|---|
| 291 | I probably would not be necessary in the cases that I have seen where unlocked access makes sense. However, I suppose one could do this for some of the potential cases, but wouldn't atomic_store_ptr be necessary also in the osd_set_reserved() function if doing so? | |
| sys/kern/kern_osd.c | ||
|---|---|---|
| 291 | It's not strictly necessary. Adding the atomic qualifier here 1) helpers readers understand that this isn't a normal, synchronized memory access, 2) provides more information to concurrency sanitizers to help avoid reporting false positives. I think it would make sense to modify osd_set_reserved() too. | |