From: Jason Gunthorpe <jgunthorpe-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
To: Haggai Eran <haggaie-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Cc: Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>,
linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
netdev-u79uwXL29TY76Z2rM5mHXA@public.gmane.org,
Liran Liss <liranl-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>,
Guy Shapiro <guysh-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>,
Shachar Raindel <raindel-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>,
Yotam Kenneth <yotamke-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>,
Matan Barak <matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Subject: Re: [PATCH v3 for-next 01/13] IB/core: Use SRCU when reading client_list or device_list
Date: Wed, 13 May 2015 09:57:37 -0600 [thread overview]
Message-ID: <20150513155737.GA16941@obsidianresearch.com> (raw)
In-Reply-To: <555306E7.9020000-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
On Wed, May 13, 2015 at 11:10:15AM +0300, Haggai Eran wrote:
> >> I guess a similar thing we can do is to rely on the context we associate
> >> with a pair of a client and a device. If such a context exist, we don't
> >> need to call client->add again. What do you think?
> >
> > I didn't look closely, isn't this enough?
> >
> > device_register:
> > mutex_lock(client_mutex);
> > down_write(devices_rwsem);
> > list_add(device_list,dev,..);
> > up_write(devices_rwsem);
> >
> > /* Caller must prevent device_register/unregister concurrancy on the
> > same dev */
> >
> > foreach(client_list)
> > .. client->add(dev,...) ..
> >
> > mutex_unlock(client_mutex)
> >
> > client_register:
> > mutex_lock(client_mutex)
> > list_add(client_list,new_client..)
> > down_read(devices_rwsem);
> > foreach(device_list)
> > .. client->add(dev,new_client,..) ..
> > up_read(devices_rwsem);
> > mutex_unlock(client_mutex)
> >
> > [Note, I didn't check this carefully, just intuitively seems like a
> > sane starting point]
>
> That could perhaps work for the RoCEv2 patch-set, as their deadlock
> relates to iterating the device list. This patch set however does an
> iteration on the client list (patch 3). Because a client remove could
> block on this iteration, you can still get a deadlock.
Really? Gross.
Still, I think you got it right in your analysis, but we don't need RCU:
device_register:
mutex_lock(modify_mutex);
down_write(lists_rwsem);
list_add(device_list,dev,..);
up_write(lists_rwsem);
// implied by modify_mutex: down_read(lists_rwsem)
foreach(client_list)
.. client->add(dev,...) ..
mutex_unlock(modify_mutex)
client_register:
mutex_lock(modify_mutex);
// implied by modify_mutex: down_read(lists_rwsem)
foreach(device_list)
.. client->add(dev,new_client...) ..
down_write(lists_rwsem);
list_add(client_list,new_client..);
up_write(lists_rwsem);
mutex_unlock(modify_mutex)
client_unregister:
mutex_lock(modify_mutex);
down_write(lists_rwsem);
list_cut(client_list,..rm_client..);
up_write(lists_rwsem);
// implied by modify_mutex: down_read(lists_rwsem)
foreach(device_list)
.. client->remove(dev,rm_client...) ..
mutex_unlock(modify_mutex)
etc. Notice the ordering.
> I think I prefer keeping the single device_mutex and the SRCU, but
> keeping the device_mutex locked as it is today, covering all of the
> registration and unregistration code. Only the new code that reads the
> client list or the device list without modification to the other list
> will use the SRCU read lock.
In this case, I don't see a justification to use RCU, we'd need a
high read load before optimizing the rwsem into RCU would make
sense.
I'm not sure you should ever use RCU until you've designed the locking
using traditional means.
Jason
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
next prev parent reply other threads:[~2015-05-13 15:57 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-05-10 10:26 [PATCH v3 for-next 00/13] Add network namespace support in the RDMA-CM Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 01/13] IB/core: Use SRCU when reading client_list or device_list Haggai Eran
2015-05-11 18:18 ` Jason Gunthorpe
2015-05-12 6:07 ` Haggai Eran
2015-05-12 18:23 ` Jason Gunthorpe
2015-05-13 8:10 ` Haggai Eran
[not found] ` <555306E7.9020000-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2015-05-13 15:57 ` Jason Gunthorpe [this message]
2015-05-14 10:35 ` Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 02/13] IB/addr: Pass network namespace as a parameter Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 05/13] IB/cm: Reference count ib_cm_ids Haggai Eran
[not found] ` <1431253604-9214-6-git-send-email-haggaie-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2015-05-11 18:34 ` Jason Gunthorpe
2015-05-12 6:50 ` Haggai Eran
[not found] ` <5551A2CB.1010407-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2015-05-12 18:54 ` Jason Gunthorpe
[not found] ` <20150512185447.GA3503-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-05-13 10:20 ` Haggai Eran
2015-05-13 16:58 ` Jason Gunthorpe
[not found] ` <20150513165823.GA20343-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-05-14 6:48 ` Haggai Eran
2015-05-15 19:11 ` Hefty, Sean
[not found] ` <1828884A29C6694DAF28B7E6B8A82373A8FDC0C3-P5GAC/sN6hkd3b2yrw5b5LfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-05-17 6:27 ` Haggai Eran
2015-05-19 19:23 ` Jason Gunthorpe
[not found] ` <20150519192353.GA23612-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-05-19 22:52 ` Hefty, Sean
[not found] ` <1828884A29C6694DAF28B7E6B8A82373A8FDD412-P5GAC/sN6hkd3b2yrw5b5LfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-05-19 23:46 ` Jason Gunthorpe
[not found] ` <20150519234654.GA26634-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-05-20 0:49 ` Hefty, Sean
2015-05-21 6:51 ` Haggai Eran
[not found] ` <555D806B.1090002-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2015-05-21 12:56 ` Hefty, Sean
[not found] ` <1431253604-9214-1-git-send-email-haggaie-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2015-05-10 10:26 ` [PATCH v3 for-next 03/13] IB/core: Find the network namespace matching connection parameters Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 04/13] IB/ipoib: Return IPoIB devices " Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 06/13] IB/cm: API to retrieve existing listening CM IDs Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 07/13] IB/cm: Expose service ID in request events Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 08/13] IB/cma: Refactor RDMA IP CM private-data parsing code Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 09/13] IB/cma: Add compare_data checks to the RDMA CM module Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 10/13] IB/cma: Separate port allocation to network namespaces Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 11/13] IB/cma: Share CM IDs between namespaces Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 12/13] IB/cma: Add support for network namespaces Haggai Eran
2015-05-10 10:26 ` [PATCH v3 for-next 13/13] IB/ucma: Take the network namespace from the process Haggai Eran
2015-05-12 17:52 ` [PATCH v3 for-next 00/13] Add network namespace support in the RDMA-CM Hefty, Sean
[not found] ` <1828884A29C6694DAF28B7E6B8A82373A8FD7B85-P5GAC/sN6hkd3b2yrw5b5LfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-05-13 8:50 ` Haggai Eran
[not found] ` <55531073.1000305-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
2015-05-13 16:45 ` Hefty, Sean
[not found] ` <1828884A29C6694DAF28B7E6B8A82373A8FDA13B-P5GAC/sN6hkd3b2yrw5b5LfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-05-13 17:18 ` Jason Gunthorpe
[not found] ` <20150513171823.GB20343-ePGOBjL8dl3ta4EC/59zMFaTQe2KTcn/@public.gmane.org>
2015-05-13 17:30 ` Hefty, Sean
[not found] ` <1828884A29C6694DAF28B7E6B8A82373A8FDA1AB-P5GAC/sN6hkd3b2yrw5b5LfspsVTdybXVpNB7YpNyf8@public.gmane.org>
2015-05-14 5:35 ` Haggai Eran
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=20150513155737.GA16941@obsidianresearch.com \
--to=jgunthorpe-epgobjl8dl3ta4ec/59zmfatqe2ktcn/@public.gmane.org \
--cc=dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org \
--cc=guysh-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org \
--cc=haggaie-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org \
--cc=linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=liranl-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org \
--cc=matanb-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org \
--cc=netdev-u79uwXL29TY76Z2rM5mHXA@public.gmane.org \
--cc=raindel-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org \
--cc=yotamke-VPRAkNaXOzVWk0Htik3J/w@public.gmane.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).