linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Mike Kravetz <mike.kravetz@oracle.com>
To: Michal Hocko <mhocko@kernel.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	Waiman Long <longman@redhat.com>,
	Matthew Wilcox <willy@infradead.org>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org,
	aneesh.kumar@linux.ibm.com
Subject: Re: [PATCH v2] mm/hugetlb: defer free_huge_page() to a workqueue
Date: Mon, 16 Dec 2019 11:08:57 -0800	[thread overview]
Message-ID: <68d466cc-2cbd-ae49-7d89-e7476c5a9c24@oracle.com> (raw)
In-Reply-To: <20191216161748.tgi2oictlfqy6azi@linux-p48b>

On 12/16/19 8:17 AM, Davidlohr Bueso wrote:
> On Mon, 16 Dec 2019, Michal Hocko wrote:
>> I am afraid that work_struct is too large to be stuffed into the struct
>> page array (because of the lockdep part).
> 
> Yeah, this needs to be done without touching struct page.
> 
> Which is why I had done the stack allocated way in this patch, but we
> cannot wait for it to complete in irq, so that's out the window. Andi
> had suggested percpu allocated work items, but having played with the
> idea over the weekend, I don't see how we can prevent another page being
> freed on the same cpu before previous work on the same cpu is complete
> (cpu0 wants to free pageA, schedules the work, in the mean time cpu0
> wants to free pageB and workerfn for pageA still hasn't been called).
> 
>> I think that it would be just safer to make hugetlb_lock irq safe. Are
>> there any other locks that would require the same?
> 
> It would be simpler. Any performance issues that arise would probably
> be only seen in microbenchmarks, assuming we want to have full irq safety.
> If we don't need to worry about hardirq, then even better.
> 
> The subpool lock would also need to be irq safe.

I do think we need to worry about hardirq.  There are no restruictions that
put_page can not be called from hardirq context. 

I am concerned about the latency of making hugetlb_lock (and potentially
subpool lock) hardirq safe.  When these locks were introduced (before my
time) the concept of making them irq safe was not considered.  Recently,
I learned that the hugetlb_lock is held for a linear scan of ALL hugetlb
pages during a cgroup reparentling operation.  That is just too long.

If there is no viable work queue solution, then I think we would like to
restructure the hugetlb locking before a change to just make hugetlb_lock
irq safe.  The idea would be to split the scope of what is done under
hugetlb_lock.  Most of it would never be executed in irq context.  Then
have a small/limited set of functionality that really needs to be irq
safe protected by an irq safe lock.

-- 
Mike Kravetz

  parent reply	other threads:[~2019-12-16 19:09 UTC|newest]

Thread overview: 27+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-12-11 19:46 [PATCH v2] hugetlbfs: Disable softIRQ when taking hugetlb_lock Waiman Long
2019-12-11 22:04 ` Mike Kravetz
2019-12-11 22:19   ` Waiman Long
2019-12-12  1:11     ` Mike Kravetz
2019-12-12  6:06       ` Davidlohr Bueso
2019-12-12  6:30         ` Davidlohr Bueso
2019-12-12 19:04           ` [PATCH v2] mm/hugetlb: defer free_huge_page() to a workqueue Davidlohr Bueso
2019-12-12 19:22             ` Mike Kravetz
2019-12-12 19:36               ` Davidlohr Bueso
2019-12-12 20:52               ` Waiman Long
2019-12-12 21:04                 ` Waiman Long
2019-12-16 13:26                 ` Michal Hocko
2019-12-16 15:38                   ` Waiman Long
2019-12-16 18:44                     ` Waiman Long
2019-12-17  9:00                       ` Michal Hocko
2019-12-17 18:05                         ` Mike Kravetz
2019-12-18 12:18                           ` hugetlbfs testing coverage (was: Re: [PATCH v2] mm/hugetlb: defer free_huge_page() to a workqueue) Michal Hocko
2019-12-12 21:01             ` [PATCH v2] mm/hugetlb: defer free_huge_page() to a workqueue Waiman Long
2019-12-16 13:37             ` Michal Hocko
2019-12-16 16:17               ` Waiman Long
2019-12-16 16:17               ` Davidlohr Bueso
2019-12-16 17:18                 ` Matthew Wilcox
2019-12-16 19:08                 ` Mike Kravetz [this message]
2019-12-16 19:13                   ` Waiman Long
2019-12-12 21:32         ` [PATCH v2] hugetlbfs: Disable softIRQ when taking hugetlb_lock Andi Kleen
2019-12-12 22:42           ` Davidlohr Bueso
2019-12-11 23:05 ` Andi Kleen

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=68d466cc-2cbd-ae49-7d89-e7476c5a9c24@oracle.com \
    --to=mike.kravetz@oracle.com \
    --cc=akpm@linux-foundation.org \
    --cc=aneesh.kumar@linux.ibm.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=longman@redhat.com \
    --cc=mhocko@kernel.org \
    --cc=willy@infradead.org \
    --subject='Re: [PATCH v2] mm/hugetlb: defer free_huge_page() to a workqueue' \
    /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

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).