From: Michael Ellerman <mpe@ellerman.id.au>
To: Nathan Lynch <nathanl@linux.ibm.com>,
Laurent Dufour <ldufour@linux.ibm.com>
Cc: tyreld@linux.ibm.com, cheloha@linux.ibm.com,
linuxppc-dev@lists.ozlabs.org
Subject: Re: [PATCH] powerpc/pseries: explicitly reschedule during drmem_lmb list traversal
Date: Fri, 31 Jul 2020 23:16:44 +1000 [thread overview]
Message-ID: <87ft974yf7.fsf@mpe.ellerman.id.au> (raw)
In-Reply-To: <875za511z6.fsf@linux.ibm.com>
Nathan Lynch <nathanl@linux.ibm.com> writes:
> Michael Ellerman <mpe@ellerman.id.au> writes:
>> Nathan Lynch <nathanl@linux.ibm.com> writes:
>>> Laurent Dufour <ldufour@linux.ibm.com> writes:
>>>> Le 28/07/2020 à 19:37, Nathan Lynch a écrit :
>>>>> The drmem lmb list can have hundreds of thousands of entries, and
>>>>> unfortunately lookups take the form of linear searches. As long as
>>>>> this is the case, traversals have the potential to monopolize the CPU
>>>>> and provoke lockup reports, workqueue stalls, and the like unless
>>>>> they explicitly yield.
>>>>>
>>>>> Rather than placing cond_resched() calls within various
>>>>> for_each_drmem_lmb() loop blocks in the code, put it in the iteration
>>>>> expression of the loop macro itself so users can't omit it.
>>>>
>>>> Is that not too much to call cond_resched() on every LMB?
>>>>
>>>> Could that be less frequent, every 10, or 100, I don't really know ?
>>>
>>> Everything done within for_each_drmem_lmb is relatively heavyweight
>>> already. E.g. calling dlpar_remove_lmb()/dlpar_add_lmb() can take dozens
>>> of milliseconds. I don't think cond_resched() is an expensive check in
>>> this context.
>>
>> Hmm, mostly.
>>
>> But there are quite a few cases like drmem_update_dt_v1():
>>
>> for_each_drmem_lmb(lmb) {
>> dr_cell->base_addr = cpu_to_be64(lmb->base_addr);
>> dr_cell->drc_index = cpu_to_be32(lmb->drc_index);
>> dr_cell->aa_index = cpu_to_be32(lmb->aa_index);
>> dr_cell->flags = cpu_to_be32(drmem_lmb_flags(lmb));
>>
>> dr_cell++;
>> }
>>
>> Which will compile to a pretty tight loop at the moment.
>>
>> Or drmem_update_dt_v2() which has two loops over all lmbs.
>>
>> And although the actual TIF check is cheap the function call to do it is
>> not free.
>>
>> So I worry this is going to make some of those long loops take even
>> longer.
>
> That's fair, and I was wrong - some of the loop bodies are relatively
> simple, not doing allocations or taking locks, etc.
>
> One way to deal is to keep for_each_drmem_lmb() as-is and add a new
> iterator that can reschedule, e.g. for_each_drmem_lmb_slow().
If we did that, how many call-sites would need converting?
Is it ~2 or ~20 or ~200?
> On the other hand... it's probably not too strong to say that the
> drmem/hotplug code is in crisis with respect to correctness and
> algorithmic complexity, so those are my overriding concerns right
> now. Yes, this change will pessimize loops that are reinitializing the
> entire drmem_lmb array on every DLPAR operation, but:
>
> 1. it doesn't make any user of for_each_drmem_lmb() less correct;
> 2. why is this code doing that in the first place, other than to
> accommodate a poor data structure choice?
>
> The duration of the system calls where this code runs are measured in
> minutes or hours on large configurations because of all the behaviors
> that are at best O(n) with the amount of memory assigned to the
> partition. For simplicity's sake I'd rather defer lower-level
> performance considerations like this until the drmem data structures'
> awful lookup properties are fixed -- hopefully in the 5.10 timeframe.
Yeah fair enough.
cheers
next prev parent reply other threads:[~2020-07-31 13:19 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-07-28 17:37 [PATCH] powerpc/pseries: explicitly reschedule during drmem_lmb list traversal Nathan Lynch
2020-07-28 17:46 ` Laurent Dufour
2020-07-28 19:19 ` Nathan Lynch
2020-07-30 0:57 ` Michael Ellerman
2020-07-30 15:01 ` Nathan Lynch
2020-07-31 13:16 ` Michael Ellerman [this message]
2020-07-31 13:52 ` Nathan Lynch
2020-08-02 12:42 ` Michael Ellerman
2020-08-10 20:03 ` Nathan Lynch
2020-08-12 1:32 ` Nathan Lynch
2020-09-09 13:27 ` Michael Ellerman
2020-09-10 7:37 ` Michael Ellerman
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=87ft974yf7.fsf@mpe.ellerman.id.au \
--to=mpe@ellerman.id.au \
--cc=cheloha@linux.ibm.com \
--cc=ldufour@linux.ibm.com \
--cc=linuxppc-dev@lists.ozlabs.org \
--cc=nathanl@linux.ibm.com \
--cc=tyreld@linux.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).