From: "André Goddard Rosa" <andre.goddard@gmail.com>
To: "Américo Wang" <xiyou.wangcong@gmail.com>
Cc: LKML <linux-kernel@vger.kernel.org>,
Andrew Morton <akpm@linux-foundation.org>,
"Serge E . Hallyn" <serue@us.ibm.com>,
Cedric Le Goater <clg@fr.ibm.com>,
Al Viro <viro@zeniv.linux.org.uk>,
Xiaotian Feng <xtfeng@gmail.com>
Subject: Re: [Patch] mqueue: fix the bad code in sys_mq_open()
Date: Thu, 25 Feb 2010 12:41:47 -0300 [thread overview]
Message-ID: <b8bf37781002250741r3836a1calec3441ced7afb863@mail.gmail.com> (raw)
In-Reply-To: <20100225134023.GB3842@hack>
Hi, Américo!
On Thu, Feb 25, 2010 at 10:40 AM, Américo Wang <xiyou.wangcong@gmail.com> wrote:
>
> Fix bad code in ipc/mqueue.c.
>
> Inspired by André Goddard Rosa's patch.
>
> a) do_create() and do_open() should not release the resources which
> are not acquired within themselves, it's their caller's work;
>
> b) Fix an fd leak;
>
> c) The goto label 'out_upsem' doesn't make any sense, rename to
> 'out_unlock';
>
> d) mntget() should be called before looking up dentry within
> ->mnt->mnt_root;
>
> e) When dealing with failure case, we should release the resources
> in a reversed order of acquiring, so reorder the goto labels;
>
> f) Remove some unused labels due to reorder.
>
> Also, this shrinks the binary by 60 bytes:
>
> text data bss dec hex filename
> 7674 1684 8 9366 2496 ipc/mqueue.o.BEFORE
> 7614 1684 8 9306 245a ipc/mqueue.o.AFTER
>
>
> Signed-off-by: WANG Cong <xiyou.wangcong@gmail.com>
> Cc: André Goddard Rosa <andre.goddard@gmail.com>
>
> ---
> diff --git a/ipc/mqueue.c b/ipc/mqueue.c
> index c79bd57..fdd09da 100644
> --- a/ipc/mqueue.c
> +++ b/ipc/mqueue.c
> @@ -650,8 +650,6 @@ static struct file *do_create(struct ipc_namespace *ipc_ns, struct dentry *dir,
> out_drop_write:
> mnt_drop_write(ipc_ns->mq_mnt);
> out:
> - dput(dentry);
> - mntput(ipc_ns->mq_mnt);
> return ERR_PTR(ret);
> }
>
> @@ -664,17 +662,11 @@ static struct file *do_open(struct ipc_namespace *ipc_ns,
> static const int oflag2acc[O_ACCMODE] = { MAY_READ, MAY_WRITE,
> MAY_READ | MAY_WRITE };
>
> - if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY)) {
> - dput(dentry);
> - mntput(ipc_ns->mq_mnt);
> + if ((oflag & O_ACCMODE) == (O_RDWR | O_WRONLY))
> return ERR_PTR(-EINVAL);
> - }
>
> - if (inode_permission(dentry->d_inode, oflag2acc[oflag & O_ACCMODE])) {
> - dput(dentry);
> - mntput(ipc_ns->mq_mnt);
> + if (inode_permission(dentry->d_inode, oflag2acc[oflag & O_ACCMODE]))
> return ERR_PTR(-EACCES);
> - }
>
> return dentry_open(dentry, ipc_ns->mq_mnt, oflag, cred);
> }
> @@ -686,7 +678,7 @@ SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, mode_t, mode,
> struct file *filp;
> char *name;
> struct mq_attr attr;
> - int fd, error;
> + int fd, error = 0;
> struct ipc_namespace *ipc_ns = current->nsproxy->ipc_ns;
>
> if (u_attr && copy_from_user(&attr, u_attr, sizeof(struct mq_attr)))
> @@ -701,13 +693,13 @@ SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, mode_t, mode,
> if (fd < 0)
> goto out_putname;
>
> + mntget(ipc_ns->mq_mnt);
> mutex_lock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
> dentry = lookup_one_len(name, ipc_ns->mq_mnt->mnt_root, strlen(name));
> if (IS_ERR(dentry)) {
> error = PTR_ERR(dentry);
> - goto out_err;
> + goto out_unlock;
> }
> - mntget(ipc_ns->mq_mnt);
>
> if (oflag & O_CREAT) {
> if (dentry->d_inode) { /* entry already exists */
> @@ -731,24 +723,23 @@ SYSCALL_DEFINE4(mq_open, const char __user *, u_name, int, oflag, mode_t, mode,
>
> if (IS_ERR(filp)) {
> error = PTR_ERR(filp);
> - goto out_putfd;
> + goto out;
> }
>
> fd_install(fd, filp);
> - goto out_upsem;
> + goto out_unlock;
>
> out:
> dput(dentry);
> - mntput(ipc_ns->mq_mnt);
> -out_putfd:
> - put_unused_fd(fd);
> -out_err:
> - fd = error;
> -out_upsem:
> +out_unlock:
> mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
> + if (error) {
> + mntput(ipc_ns->mq_mnt);
> + put_unused_fd(fd);
> + }
> out_putname:
> putname(name);
> - return fd;
> + return error;
> }
>
> SYSCALL_DEFINE1(mq_unlink, const char __user *, u_name)
I have some questions below:
Inside do_open() and do_create() on mqueue.c, we call
dentry_open()/__dentry_open().
If dentry_open() fails, it'll automatically call:
dput(dentry);
mntput(mnt);
Then we'll go here and set "error":
> if (IS_ERR(filp)) {
> error = PTR_ERR(filp);
> - goto out_putfd;
> + goto out;
> }
>
And finally here:
> out:
> dput(dentry);
> - mntput(ipc_ns->mq_mnt);
> -out_putfd:
> - put_unused_fd(fd);
> -out_err:
> - fd = error;
> -out_upsem:
> +out_unlock:
> mutex_unlock(&ipc_ns->mq_mnt->mnt_root->d_inode->i_mutex);
> + if (error) {
> + mntput(ipc_ns->mq_mnt);
> + put_unused_fd(fd);
> + }
Is it ok to call:
dput(dentry);
mntput(ipc_ns->mq_mnt);
multiple times?
I also do not see where you set "error" to "fd":
> - return fd;
> + return error;
Am I missing something?
Could you please base your patch on top of mine?
Thank you,
André
next prev parent reply other threads:[~2010-02-25 15:42 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-02-23 7:04 [PATCH 0/6] Fix file descriptor leak on user-space processes and cleanup André Goddard Rosa
2010-02-23 7:04 ` [PATCH 1/6] mqueue: remove unneeded info->messages initialization André Goddard Rosa
2010-02-23 7:04 ` [PATCH 2/6] mqueue: apply mathematics distributivity on mq_bytes calculation André Goddard Rosa
2010-02-23 7:04 ` [PATCH 3/6] mqueue: simplify do_open() error handling André Goddard Rosa
2010-02-23 7:04 ` [PATCH 4/6] mqueue: only set error codes if they are really necessary André Goddard Rosa
2010-02-23 7:04 ` [PATCH 5/6] mqueue: fix typo "failues" -> "failures" André Goddard Rosa
2010-02-23 7:04 ` [PATCH 6/6] mqueue: fix mq_open() file descriptor leak on user-space processes André Goddard Rosa
2010-02-25 3:35 ` Américo Wang
2010-02-25 4:00 ` Xiaotian Feng
2010-02-25 4:25 ` Américo Wang
2010-02-25 6:59 ` Américo Wang
2010-02-25 10:49 ` Xiaotian Feng
2010-02-25 13:17 ` Américo Wang
2010-02-25 13:40 ` [Patch] mqueue: fix the bad code in sys_mq_open() Américo Wang
2010-02-25 15:41 ` André Goddard Rosa [this message]
2010-02-25 16:15 ` Américo Wang
2010-03-03 19:54 ` Al Viro
2010-02-25 10:56 ` [PATCH 6/6] mqueue: fix mq_open() file descriptor leak on user-space processes Xiaotian Feng
2010-02-24 22:01 ` [PATCH 0/6] Fix file descriptor leak on user-space processes and cleanup Andrew Morton
2010-02-25 15:52 ` André Goddard Rosa
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=b8bf37781002250741r3836a1calec3441ced7afb863@mail.gmail.com \
--to=andre.goddard@gmail.com \
--cc=akpm@linux-foundation.org \
--cc=clg@fr.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=serue@us.ibm.com \
--cc=viro@zeniv.linux.org.uk \
--cc=xiyou.wangcong@gmail.com \
--cc=xtfeng@gmail.com \
/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®