Skip to content

blk/aio: fix long batch (64+K entries) submission.#56352

Merged
yuriw merged 6 commits intoceph:mainfrom
ifed01:wip-ifed-many-many-extents-read
Jul 18, 2024
Merged

blk/aio: fix long batch (64+K entries) submission.#56352
yuriw merged 6 commits intoceph:mainfrom
ifed01:wip-ifed-many-many-extents-read

Conversation

@ifed01
Copy link
Contributor

@ifed01 ifed01 commented Mar 20, 2024

This is a revival of #44065
Includes some fixes and UT.

Fixes: https://tracker.ceph.com/issues/64966

Contribution Guidelines

  • To sign and title your commits, please refer to Submitting Patches to Ceph.

  • If you are submitting a fix for a stable branch (e.g. "quincy"), please refer to Submitting Patches to Ceph - Backports for the proper workflow.

  • When filling out the below checklist, you may click boxes directly in the GitHub web UI. When entering or editing the entire PR message in the GitHub web UI editor, you may also select a checklist item by adding an x between the brackets: [x]. Spaces and capitalization matter when checking off items this way.

Checklist

  • Tracker (select at least one)
    • References tracker ticket
    • Very recent bug; references commit where it was introduced
    • New feature (ticket optional)
    • Doc update (no ticket needed)
    • Code cleanup (no ticket needed)
  • Component impact
    • Affects Dashboard, opened tracker ticket
    • Affects Orchestrator, opened tracker ticket
    • No impact that needs to be tracked
  • Documentation (select at least one)
    • Updates relevant documentation
    • No doc update is appropriate
  • Tests (select at least one)
Show available Jenkins commands
  • jenkins retest this please
  • jenkins test classic perf
  • jenkins test crimson perf
  • jenkins test signed
  • jenkins test make check
  • jenkins test make check arm64
  • jenkins test submodules
  • jenkins test dashboard
  • jenkins test dashboard cephadm
  • jenkins test api
  • jenkins test docs
  • jenkins render docs
  • jenkins test ceph-volume all
  • jenkins test ceph-volume tox
  • jenkins test windows
  • jenkins test rook e2e

@ifed01 ifed01 requested a review from a team as a code owner March 20, 2024 21:06
@ifed01 ifed01 added the bug-fix label Mar 20, 2024
@ifed01 ifed01 requested review from aclamk and rzarzynski March 20, 2024 21:07
@ifed01 ifed01 force-pushed the wip-ifed-many-many-extents-read branch from c29dd7b to bc78b3b Compare March 20, 2024 21:11
@ifed01
Copy link
Contributor Author

ifed01 commented Mar 21, 2024

jenkins test make check

@ifed01
Copy link
Contributor Author

ifed01 commented Mar 21, 2024

jenkins test api

int r;

aio_iter cur = begin;
struct aio_t *piocb[aios_size];
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The problem, as I understand it, is very simple: aios_size got an overflow.
I think solution should be to either:

  1. change uint16_t to uint32_t (preferred)
  2. count elements between begin and end; use alloca to allocate piocb.

The only benefit of current solution is that it does minimize stack allocation, but at cost of added complexity. I do not think its worth it.
Note that for every entry in piocb we will be reading at least 4K from the drive. I think we can afford having entire piocb on stack.

Copy link
Contributor Author

@ifed01 ifed01 Apr 15, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@aclamk - you might be missing one more aspect - amount of chunks (and hence amount IOs issued in a single batch) in an object is determined (to some degree) by cluster's users. So it becomes possible for them to use "bad" objects which could DDOS OSDs or do something inappropriate.
And this patch provides a sort of protection against that - at aio level at least..

Given that I'd prefer not to [ab]use stack for this task.

ifed01 and others added 3 commits April 12, 2024 14:35
Signed-off-by: Igor Fedotov <igor.fedotov@croit.io>
Signed-off-by: Igor Fedotov <igor.fedotov@croit.io>
Due to aios_size being a uint16 and the source value for the actual
call being an int there was a possible overflow. This was "fixed"
with an assert, however that still causes a crash.

This commit removes the need for aios_size completely by iterating
over the list and submitting it in max_iodepth batches.

Fixes: https://tracker.ceph.com/issues/46366
Signed-off-by: Robin Geuze <robin.geuze@nl.team.blue>
(cherry picked from commit f87db49)
@ifed01 ifed01 force-pushed the wip-ifed-many-many-extents-read branch 2 times, most recently from 444a0ac to 3666d9d Compare April 15, 2024 13:51
@ifed01
Copy link
Contributor Author

ifed01 commented Apr 16, 2024

jenkins test api

cur->aio.aiocb.aio_sigevent.sigev_notify = SIGEV_KEVENT;
cur->aio.aiocb.aio_sigevent.sigev_notify_kqueue = ctx;
cur->aio.aiocb.aio_sigevent.sigev_value.sival_ptr = &(*cur);
r = aio_read(&cur->aio.aiocb);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

'aio_read'? How on earth POSIXAIO could have ever worked with ceph? I refuse to believe that there were never single aio request to write one chunk.
Proposal:
Lets replace implementation here with #error Implementation removed..

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah, that's funny. Replaced with aio_write()

r = io_submit(ctx, toSubmit, (struct iocb**)(piocb + itemCount0));
if (r >= 0 && r < toSubmit) {
r = -EAGAIN;
itemCount0 += (r > 0 ? r : 0);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Easy task for an optimizer: itemCount0 += 0;, as r==-EAGAIN :)

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I believed I've fixed that... Let me check.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

missed the push, should be fine now

@ifed01 ifed01 force-pushed the wip-ifed-many-many-extents-read branch 2 times, most recently from 41b2fa9 to 4790da3 Compare April 18, 2024 10:58
ifed01 and others added 3 commits April 18, 2024 14:08
Signed-off-by: Igor Fedotov <igor.fedotov@croit.io>
Signed-off-by: Adam Kupczyk <akupczyk@ibm.com>
Signed-off-by: Igor Fedotov <igor.fedotov@croit.io>
@ifed01 ifed01 force-pushed the wip-ifed-many-many-extents-read branch from 4790da3 to acf7f15 Compare April 18, 2024 11:08
@aclamk
Copy link
Contributor

aclamk commented Apr 30, 2024

jenkins test make check

@ifed01
Copy link
Contributor Author

ifed01 commented May 1, 2024

jenkins test make check

@yuriw
Copy link
Contributor

yuriw commented May 3, 2024

This PR is under test in https://tracker.ceph.com/issues/65796.

@github-actions
Copy link

github-actions bot commented Jul 2, 2024

This pull request has been automatically marked as stale because it has not had any activity for 60 days. It will be closed if no further activity occurs for another 30 days.
If you are a maintainer or core committer, please follow-up on this pull request to identify what steps should be taken by the author to move this proposed change forward.
If you are the author of this pull request, thank you for your proposed contribution. If you believe this change is still appropriate, please ensure that any feedback has been addressed and ask for a code review.

@github-actions github-actions bot added the stale label Jul 2, 2024
@aclamk aclamk removed the stale label Jul 9, 2024
@ifed01
Copy link
Contributor Author

ifed01 commented Jul 11, 2024

unstale

@kamoltat
Copy link
Member

@yuriw yuriw merged commit 28955bd into ceph:main Jul 18, 2024
@ifed01 ifed01 deleted the wip-ifed-many-many-extents-read branch July 18, 2024 21:17
NitzanMordhai pushed a commit to NitzanMordhai/ceph that referenced this pull request Aug 1, 2024
…read

blk/aio: fix long batch (64+K entries) submission.

Reviewed-by: Adam Kupczyk <akupczyk@redhat.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

6 participants