From: Boris Ostrovsky <boris.ostrovsky@oracle.com>
To: Stefano Stabellini <sstabellini@kernel.org>
Cc: xen-devel@lists.xen.org, linux-kernel@vger.kernel.org,
jgross@suse.com, Stefano Stabellini <stefano@aporeto.com>
Subject: Re: [PATCH v6 12/13] xen/pvcalls: implement release command
Date: Thu, 26 Oct 2017 10:11:03 -0400 [thread overview]
Message-ID: <daae2a95-b1c0-1d19-9a65-552801789ac2@oracle.com> (raw)
In-Reply-To: <alpine.DEB.2.10.1710251557200.574@sstabellini-ThinkPad-X260>
On 10/25/2017 07:00 PM, Stefano Stabellini wrote:
> On Wed, 25 Oct 2017, Boris Ostrovsky wrote:
>> On 10/24/2017 01:33 PM, Stefano Stabellini wrote:
>>> Send PVCALLS_RELEASE to the backend and wait for a reply. Take both
>>> in_mutex and out_mutex to avoid concurrent accesses. Then, free the
>>> socket.
>>>
>>> For passive sockets, check whether we have already pre-allocated an
>>> active socket for the purpose of being accepted. If so, free that as
>>> well.
>>>
>>> Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
>>> CC: boris.ostrovsky@oracle.com
>>> CC: jgross@suse.com
>>> ---
>>> drivers/xen/pvcalls-front.c | 100 ++++++++++++++++++++++++++++++++++++++++++++
>>> drivers/xen/pvcalls-front.h | 1 +
>>> 2 files changed, 101 insertions(+)
>>>
>>> diff --git a/drivers/xen/pvcalls-front.c b/drivers/xen/pvcalls-front.c
>>> index 4a413ff..7abc039 100644
>>> --- a/drivers/xen/pvcalls-front.c
>>> +++ b/drivers/xen/pvcalls-front.c
>>> @@ -199,6 +199,23 @@ static irqreturn_t pvcalls_front_event_handler(int irq, void *dev_id)
>>> static void pvcalls_front_free_map(struct pvcalls_bedata *bedata,
>>> struct sock_mapping *map, bool locked)
>>> {
>>> + int i;
>>> +
>>> + unbind_from_irqhandler(map->active.irq, map);
>>> +
>>> + if (!locked)
>>> + spin_lock(&bedata->socket_lock);
>>> + if (!list_empty(&map->list))
>>> + list_del_init(&map->list);
>>> + if (!locked)
>>> + spin_unlock(&bedata->socket_lock);
>>> +
>>> + for (i = 0; i < (1 << PVCALLS_RING_ORDER); i++)
>>> + gnttab_end_foreign_access(map->active.ring->ref[i], 0, 0);
>>> + gnttab_end_foreign_access(map->active.ref, 0, 0);
>>> + free_page((unsigned long)map->active.ring);
>>> +
>>> + kfree(map);
>>> }
>>>
>>> static irqreturn_t pvcalls_front_conn_handler(int irq, void *sock_map)
>>> @@ -966,6 +983,89 @@ unsigned int pvcalls_front_poll(struct file *file, struct socket *sock,
>>> return ret;
>>> }
>>>
>>> +int pvcalls_front_release(struct socket *sock)
>>> +{
>>> + struct pvcalls_bedata *bedata;
>>> + struct sock_mapping *map;
>>> + int req_id, notify, ret;
>>> + struct xen_pvcalls_request *req;
>>> +
>> ..
>>
>>> +
>>> + if (map->active_socket) {
>>> + /*
>>> + * Set in_error and wake up inflight_conn_req to force
>>> + * recvmsg waiters to exit.
>>> + */
>>> + map->active.ring->in_error = -EBADF;
>>> + wake_up_interruptible(&map->active.inflight_conn_req);
>>> +
>>> + /*
>>> + * Wait until there are no more waiters on the mutexes.
>>> + * We know that no new waiters can be added because sk_send_head
>>> + * is set to NULL -- we only need to wait for the existing
>>> + * waiters to return.
>>> + */
>>> + while (!mutex_trylock(&map->active.in_mutex) ||
>>> + !mutex_trylock(&map->active.out_mutex))
>>> + cpu_relax();
>>> +
>>> + pvcalls_front_free_map(bedata, map, false);
>>> + } else {
>>> + spin_lock(&bedata->socket_lock);
>>> + if (READ_ONCE(map->passive.inflight_req_id) !=
>>> + PVCALLS_INVALID_ID) {
>>> + pvcalls_front_free_map(bedata,
>>> + map->passive.accept_map, true);
>>> + }
>>> + list_del(&map->list);
>>> + kfree(map);
>>> + spin_unlock(&bedata->socket_lock);
>> We have different locking rules in pvcalls_front_free_map() for each of
>> those clauses in that in the first case we are doing grant table
>> operations and free_page() without the lock and in the second case we
>> are holding it. Is it possible to restructure this so that we prune the
>> lists under the lock (possibly in this routine) and call
>> pvcalls_front_free_map() lock-less?
> Yes, it is possible. However, pvcalls_front_free_map is called from a
> couple of other places (pvcalls_front_accept and pvcalls_front_remove)
> and we would have to add the code to remove the map from the list there
> as well. I am not sure it is worth it.
>
> I don't have a strong opinion on this. Let me know which way you prefer.
I didn't realize this is called from multiple places.
How about
spin_lock(&bedata->socket_lock);
if (READ_ONCE(map->passive.inflight_req_id) !=
PVCALLS_INVALID_ID) {
if (!list_empty(&map->passive.accept_map->list))
list_del(&map->passive.accept_map->list);
}
list_del(&map->list);
spin_unlock(&bedata->socket_lock);
pvcalls_front_free_map(bedata, map->passive.accept_map, true);
kfree(map);
(I may have messed up list pointers here)
This would be slightly inefficient in that you drop the lock and then grab it again (only to find that the list is empty, presumably) but it makes pvcalls_front_free_map()'s behavior more consistent wrt locking. (Or maybe you will need to pass a bool to pvcalls_front_free_map() to indicate whether the map needs to be removed from the list for other call sites, in which case it will replace the 'locked' argument)
-boris
next prev parent reply other threads:[~2017-10-26 14:10 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-10-24 17:33 [PATCH v6 00/13] introduce the Xen PV Calls frontend Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 01/13] xen/pvcalls: introduce the pvcalls xenbus frontend Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 02/13] xen/pvcalls: implement frontend disconnect Stefano Stabellini
2017-10-25 13:49 ` Boris Ostrovsky
2017-10-25 23:22 ` Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 03/13] xen/pvcalls: connect to the backend Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 04/13] xen/pvcalls: implement socket command and handle events Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 05/13] xen/pvcalls: implement connect command Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 06/13] xen/pvcalls: implement bind command Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 07/13] xen/pvcalls: implement listen command Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 08/13] xen/pvcalls: implement accept command Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 09/13] xen/pvcalls: implement sendmsg Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 10/13] xen/pvcalls: implement recvmsg Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 11/13] xen/pvcalls: implement poll command Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 12/13] xen/pvcalls: implement release command Stefano Stabellini
2017-10-25 14:32 ` Boris Ostrovsky
2017-10-25 23:00 ` Stefano Stabellini
2017-10-26 14:11 ` Boris Ostrovsky [this message]
2017-10-26 18:21 ` Stefano Stabellini
2017-10-24 17:33 ` [PATCH v6 13/13] xen: introduce a Kconfig option to enable the pvcalls frontend Stefano Stabellini
2017-10-25 14:39 ` Boris Ostrovsky
2017-10-25 22:56 ` Stefano Stabellini
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=daae2a95-b1c0-1d19-9a65-552801789ac2@oracle.com \
--to=boris.ostrovsky@oracle.com \
--cc=jgross@suse.com \
--cc=linux-kernel@vger.kernel.org \
--cc=sstabellini@kernel.org \
--cc=stefano@aporeto.com \
--cc=xen-devel@lists.xen.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
all inboxes | Powered by JetHome®