From: Davidlohr Bueso <dave@stgolabs.net>
To: jgg@ziepe.ca, akpm@linux-foundation.org
Cc: dledford@redhat.com, jgg@mellanox.com, jack@suse.cz,
willy@infradead.org, ira.weiny@intel.com,
linux-rdma@vger.kernel.org, linux-mm@kvack.org,
linux-kernel@vger.kernel.org, dave@stgolabs.net,
Davidlohr Bueso <dbueso@suse.de>
Subject: [PATCH 6/6] drivers/IB,core: reduce scope of mmap_sem
Date: Wed, 6 Feb 2019 09:59:20 -0800 [thread overview]
Message-ID: <20190206175920.31082-7-dave@stgolabs.net> (raw)
In-Reply-To: <20190206175920.31082-1-dave@stgolabs.net>
ib_umem_get() uses gup_longterm() and relies on the lock to
stabilze the vma_list, so we cannot really get rid of mmap_sem
altogether, but now that the counter is atomic, we can get of
some complexity that mmap_sem brings with only pinned_vm.
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Signed-off-by: Davidlohr Bueso <dbueso@suse.de>
---
drivers/infiniband/core/umem.c | 41 ++---------------------------------------
1 file changed, 2 insertions(+), 39 deletions(-)
diff --git a/drivers/infiniband/core/umem.c b/drivers/infiniband/core/umem.c
index 678abe1afcba..b69d3efa8712 100644
--- a/drivers/infiniband/core/umem.c
+++ b/drivers/infiniband/core/umem.c
@@ -165,15 +165,12 @@ struct ib_umem *ib_umem_get(struct ib_udata *udata, unsigned long addr,
lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT;
- down_write(&mm->mmap_sem);
- new_pinned = atomic64_read(&mm->pinned_vm) + npages;
+ new_pinned = atomic64_add_return(npages, &mm->pinned_vm);
if (new_pinned > lock_limit && !capable(CAP_IPC_LOCK)) {
- up_write(&mm->mmap_sem);
+ atomic64_sub(npages, &mm->pinned_vm);
ret = -ENOMEM;
goto out;
}
- atomic64_set(&mm->pinned_vm, new_pinned);
- up_write(&mm->mmap_sem);
cur_base = addr & PAGE_MASK;
@@ -233,9 +230,7 @@ struct ib_umem *ib_umem_get(struct ib_udata *udata, unsigned long addr,
umem_release:
__ib_umem_release(context->device, umem, 0);
vma:
- down_write(&mm->mmap_sem);
atomic64_sub(ib_umem_num_pages(umem), &mm->pinned_vm);
- up_write(&mm->mmap_sem);
out:
if (vma_list)
free_page((unsigned long) vma_list);
@@ -258,25 +253,12 @@ static void __ib_umem_release_tail(struct ib_umem *umem)
kfree(umem);
}
-static void ib_umem_release_defer(struct work_struct *work)
-{
- struct ib_umem *umem = container_of(work, struct ib_umem, work);
-
- down_write(&umem->owning_mm->mmap_sem);
- atomic64_sub(ib_umem_num_pages(umem), &umem->owning_mm->pinned_vm);
- up_write(&umem->owning_mm->mmap_sem);
-
- __ib_umem_release_tail(umem);
-}
-
/**
* ib_umem_release - release memory pinned with ib_umem_get
* @umem: umem struct to release
*/
void ib_umem_release(struct ib_umem *umem)
{
- struct ib_ucontext *context = umem->context;
-
if (umem->is_odp) {
ib_umem_odp_release(to_ib_umem_odp(umem));
__ib_umem_release_tail(umem);
@@ -285,26 +267,7 @@ void ib_umem_release(struct ib_umem *umem)
__ib_umem_release(umem->context->device, umem, 1);
- /*
- * We may be called with the mm's mmap_sem already held. This
- * can happen when a userspace munmap() is the call that drops
- * the last reference to our file and calls our release
- * method. If there are memory regions to destroy, we'll end
- * up here and not be able to take the mmap_sem. In that case
- * we defer the vm_locked accounting a workqueue.
- */
- if (context->closing) {
- if (!down_write_trylock(&umem->owning_mm->mmap_sem)) {
- INIT_WORK(&umem->work, ib_umem_release_defer);
- queue_work(ib_wq, &umem->work);
- return;
- }
- } else {
- down_write(&umem->owning_mm->mmap_sem);
- }
atomic64_sub(ib_umem_num_pages(umem), &umem->owning_mm->pinned_vm);
- up_write(&umem->owning_mm->mmap_sem);
-
__ib_umem_release_tail(umem);
}
EXPORT_SYMBOL(ib_umem_release);
--
2.16.4
next prev parent reply other threads:[~2019-02-06 18:00 UTC|newest]
Thread overview: 14+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-02-06 17:59 [PATCH v3 0/6] mm: make pinned_vm atomic and simplify users Davidlohr Bueso
2019-02-06 17:59 ` [PATCH 1/6] mm: make mm->pinned_vm an atomic64 counter Davidlohr Bueso
2019-02-06 17:59 ` [PATCH 2/6] drivers/mic/scif: do not use mmap_sem Davidlohr Bueso
2019-02-06 17:59 ` [PATCH 3/6] drivers/IB,qib: optimize mmap_sem usage Davidlohr Bueso
2019-02-06 17:59 ` [PATCH 4/6] drivers/IB,hfi1: do not se mmap_sem Davidlohr Bueso
2019-02-06 17:59 ` [PATCH 5/6] drivers/IB,usnic: reduce scope of mmap_sem Davidlohr Bueso
2019-02-06 17:59 ` Davidlohr Bueso [this message]
2019-02-07 1:31 ` [PATCH 7/6] Documentation/infiniband: update from locked to pinned_vm Davidlohr Bueso
2019-02-12 0:35 ` Weiny, Ira
2019-02-07 20:12 ` [PATCH v3 0/6] mm: make pinned_vm atomic and simplify users Jason Gunthorpe
-- strict thread matches above, loose matches on Subject: below --
2019-01-21 17:42 [PATCH v2 -next " Davidlohr Bueso
2019-01-21 17:42 ` [PATCH 6/6] drivers/IB,core: reduce scope of mmap_sem Davidlohr Bueso
2019-01-21 18:32 ` Jason Gunthorpe
2019-01-21 19:12 ` Davidlohr Bueso
2019-01-21 21:53 ` Christopher Lameter
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=20190206175920.31082-7-dave@stgolabs.net \
--to=dave@stgolabs.net \
--cc=akpm@linux-foundation.org \
--cc=dbueso@suse.de \
--cc=dledford@redhat.com \
--cc=ira.weiny@intel.com \
--cc=jack@suse.cz \
--cc=jgg@mellanox.com \
--cc=jgg@ziepe.ca \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux-rdma@vger.kernel.org \
--cc=willy@infradead.org \
/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).