This repository has been archived by the owner on Feb 26, 2020. It is now read-only.
Changes needed by newer Illumos code #300
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
These kstat interfaces are required to port "Illumos #3537 want pool io kstats" to ZFS on Linux. kstat_waitq_enter() kstat_waitq_exit() kstat_runq_enter() kstat_runq_exit() Additionally, zero out the ks_data buffer in __kstat_create() so that the kstat_io_t counters are initialized to zero.
While porting Illumos #3537 I found that ks_lock member of kstat_t structure is different between Illumos and SPL. It is a pointer to the kmutex_t in Illumos, but the mutex lock itself in SPL. Apparently Illumos kstat API allows consumer to override the lock if required. With SPL implementation it is not possible anymore. Things were alright until the first attempt to actually override the lock. Porting of Illumos #3537 introduced such code for the first time. In order to provide the Solaris/Illumos like functionality we: 1. convert ks_lock to "kmutex_t *ks_lock" 2. create a new field "kmutex_t ks_private_lock" 3. On kstat_create() ks_lock = &ks_private_lock Thus if consumer doesn't care we still have our internal lock in use. If, however, consumer does care she has a chance to set ks_lock to anything else before calling kstat_install(). The rest of the code will use ks_lock regardless of its origin.
Needed for Illumos #3852. This interface is supposed to support a
variable-resolution timeout with nanosecond granularity. This
implementation just rounds up to jiffie resolution, as this was the most
expedient solution, and nanosecond resolution is rarely needed for
real-world performance tuning. Add flags from sys/callo.h as these are
used to control the behavior of cv_timedwait_hires(). Specifically,
CALLOUT_FLAG_ABSOLUTE
Normally, the expiration passed to the timeout API functions is
an expiration interval. If this flag is specified, then it is
interpreted as the expiration time itself.
CALLOUT_FLAG_ROUNDUP
Roundup the expiration time to the next resolution boundary. If this
flag is not specified, the expiration time is rounded down.
References:
https://www.illumos.org/issues/3582
illumos/illumos-gate@0689f76
This is needed for the Illumos #4045 write throttle patch. It is used in the arc eviction code to avoid blocking all arc activity by sitting on arcs_mtx too long.
The Open Solaris man page states that KM_NOSLEEP will "Return NULL immediately if memory is not available". However, we currently do not honor this behavior. We map KM_NOSLEEP to a Linux equivalent that contains __GFP_NORETRY, which is meant to prevent this behavior. We current this by checking for __GFP_NORETRY in spl_kmem_cache_alloc() and spl_cache_refill() If it is set, exit the busy loop and return with whatever we have done. Signed-off-by: Richard Yao <ryao@gentoo.org>
The SPL's SLAB code was written with an optimization to preconstruct objects before they are used. This is semantically differs from the Solaris behavior where objects are constructed and deconstructed on demand. A consequence of preallocation is that buffers that should be zeroed by constructors will contain stale data whenever the buffer is reused. I discovered this when experimenting with the conversion of important ARC memory allocations to SLAB allocation. We switch to on-demand construction and deconstruction to correct this. Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Richard Yao <ryao@gentoo.org>
|
Most of these changes exist in other pull requests which I hope to get merged fairly soon. |
|
@ryao Several of these changes have been merged, can you refresh this branch against master when you get a chance. |
|
The only changes remaining unmerged in this pull request relate to memory management. It would be best to open a new pull request with those patches reworked to include the feedback above. |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
These are all of the commits needed to support newer Illumos code. It also supersedes #286. I am opening this pull request for comments.