From: Al Viro <viro@ftp.linux.org.uk>
To: Sergey Vlasov <vsu@altlinux.ru>
Cc: Kay Sievers <kay.sievers@vrfy.org>,
Roderich.Schupp.extern@mch.siemens.de,
linux-kernel@vger.kernel.org,
linux-hotplug-devel@lists.sourceforge.net,
Greg KH <greg@kroah.com>
Subject: Re: Race between "mount" uevent and /proc/mounts?
Date: Thu, 03 Nov 2005 08:07:13 +0000 [thread overview]
Message-ID: <20051103080713.GD7992@ftp.linux.org.uk> (raw)
In-Reply-To: <20051102130118.GA23142@master.mivlgu.local>
On Wed, Nov 02, 2005 at 04:01:18PM +0300, Sergey Vlasov wrote:
> @@ -120,6 +122,10 @@ static void detach_mnt(struct vfsmount *
> list_del_init(&mnt->mnt_child);
> list_del_init(&mnt->mnt_hash);
> old_nd->dentry->d_mounted--;
> + if (current->namespace) {
> + current->namespace->event++;
> + wake_up_interruptible(&mounts_wait);
> + }
> }
>
> static void attach_mnt(struct vfsmount *mnt, struct nameidata *nd)
Ugh... So umount -l gives one hell of a spew for no good reason.
> @@ -129,6 +135,8 @@ static void attach_mnt(struct vfsmount *
> list_add(&mnt->mnt_hash, mount_hashtable+hash(nd->mnt, nd->dentry));
> list_add_tail(&mnt->mnt_child, &nd->mnt->mnt_mounts);
> nd->dentry->d_mounted++;
> + current->namespace->event++;
> + wake_up_interruptible(&mounts_wait);
> }
Bad idea - copy_tree() will spew *and* we get bogus events on CLONE_NEWNS
(i.e. current->namespace is not even the namespace being modified).
> @@ -1093,6 +1104,7 @@ int copy_namespace(int flags, struct tas
> atomic_set(&new_ns->count, 1);
> init_rwsem(&new_ns->sem);
> INIT_LIST_HEAD(&new_ns->list);
> + new_ns->event = 0;
BTW, I'd rather make that queue per-namespace...
> + down_read(&namespace->sem);
> + if (private->last_event != namespace->event) {
> + private->last_event = namespace->event;
> + ret = POLLIN | POLLRDNORM;
Umm... I'd rather use POLLERR, since POLLIN doesn't apply here - it's not
a stream of data that gives blocking read() when reached the end.
IMO the right approach is to have global event counter and do the following:
if (namespace->event != event) {
namespace->event = event;
wake...
}
in tree modifications and bump the event counter as soon as we decide
to do something. I'll see how well that works with shared-subtree stuff
and post the patch if it turns out to be usable...
-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
Linux-hotplug-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel
WARNING: multiple messages have this Message-ID (diff)
From: Al Viro <viro@ftp.linux.org.uk>
To: Sergey Vlasov <vsu@altlinux.ru>
Cc: Kay Sievers <kay.sievers@vrfy.org>,
Roderich.Schupp.extern@mch.siemens.de,
linux-kernel@vger.kernel.org,
linux-hotplug-devel@lists.sourceforge.net,
Greg KH <greg@kroah.com>
Subject: Re: Race between "mount" uevent and /proc/mounts?
Date: Thu, 3 Nov 2005 08:07:13 +0000 [thread overview]
Message-ID: <20051103080713.GD7992@ftp.linux.org.uk> (raw)
In-Reply-To: <20051102130118.GA23142@master.mivlgu.local>
On Wed, Nov 02, 2005 at 04:01:18PM +0300, Sergey Vlasov wrote:
> @@ -120,6 +122,10 @@ static void detach_mnt(struct vfsmount *
> list_del_init(&mnt->mnt_child);
> list_del_init(&mnt->mnt_hash);
> old_nd->dentry->d_mounted--;
> + if (current->namespace) {
> + current->namespace->event++;
> + wake_up_interruptible(&mounts_wait);
> + }
> }
>
> static void attach_mnt(struct vfsmount *mnt, struct nameidata *nd)
Ugh... So umount -l gives one hell of a spew for no good reason.
> @@ -129,6 +135,8 @@ static void attach_mnt(struct vfsmount *
> list_add(&mnt->mnt_hash, mount_hashtable+hash(nd->mnt, nd->dentry));
> list_add_tail(&mnt->mnt_child, &nd->mnt->mnt_mounts);
> nd->dentry->d_mounted++;
> + current->namespace->event++;
> + wake_up_interruptible(&mounts_wait);
> }
Bad idea - copy_tree() will spew *and* we get bogus events on CLONE_NEWNS
(i.e. current->namespace is not even the namespace being modified).
> @@ -1093,6 +1104,7 @@ int copy_namespace(int flags, struct tas
> atomic_set(&new_ns->count, 1);
> init_rwsem(&new_ns->sem);
> INIT_LIST_HEAD(&new_ns->list);
> + new_ns->event = 0;
BTW, I'd rather make that queue per-namespace...
> + down_read(&namespace->sem);
> + if (private->last_event != namespace->event) {
> + private->last_event = namespace->event;
> + ret = POLLIN | POLLRDNORM;
Umm... I'd rather use POLLERR, since POLLIN doesn't apply here - it's not
a stream of data that gives blocking read() when reached the end.
IMO the right approach is to have global event counter and do the following:
if (namespace->event != event) {
namespace->event = event;
wake...
}
in tree modifications and bump the event counter as soon as we decide
to do something. I'll see how well that works with shared-subtree stuff
and post the patch if it turns out to be usable...
next prev parent reply other threads:[~2005-11-03 8:07 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-10-25 13:20 Race between "mount" uevent and /proc/mounts? Schupp Roderich (extern) BenQ MD PD SWP 2 CM MCH
2005-10-25 14:00 ` Al Viro
2005-10-26 10:27 ` Sergey Vlasov
2005-10-26 10:27 ` Sergey Vlasov
2005-10-26 11:15 ` Al Viro
2005-10-26 11:15 ` Al Viro
2005-10-26 14:34 ` Kay Sievers
2005-10-26 14:34 ` Kay Sievers
2005-10-26 14:45 ` Xavier Bestel
2005-10-26 14:45 ` Xavier Bestel
2005-10-26 19:28 ` Al Viro
2005-10-26 19:28 ` Al Viro
2005-11-01 0:28 ` Kay Sievers
2005-11-01 0:28 ` Kay Sievers
2005-11-01 3:58 ` Kay Sievers
2005-11-01 3:58 ` Kay Sievers
2005-11-01 19:54 ` Sergey Vlasov
2005-11-01 19:54 ` Sergey Vlasov
2005-11-01 21:35 ` Kay Sievers
2005-11-01 21:35 ` Kay Sievers
2005-11-02 13:01 ` Sergey Vlasov
2005-11-02 13:01 ` Sergey Vlasov
2005-11-03 8:07 ` Al Viro [this message]
2005-11-03 8:07 ` Al Viro
2005-11-03 10:52 ` Sergey Vlasov
2005-11-03 10:52 ` Sergey Vlasov
2005-11-03 11:30 ` Al Viro
2005-11-03 11:30 ` Al Viro
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=20051103080713.GD7992@ftp.linux.org.uk \
--to=viro@ftp.linux.org.uk \
--cc=Roderich.Schupp.extern@mch.siemens.de \
--cc=greg@kroah.com \
--cc=kay.sievers@vrfy.org \
--cc=linux-hotplug-devel@lists.sourceforge.net \
--cc=linux-kernel@vger.kernel.org \
--cc=vsu@altlinux.ru \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.