[PATCH] cifs: don't leak -EAGAIN for stat() during reconnect

[PATCH] cifs: don't leak -EAGAIN for stat() during reconnect

[Ronnie Sahlberg]

If from cifs_revalidate_dentry_attr() the SMB2/QUERY_INFO call fails with an
error, such as STATUS_SESSION_EXPIRED, causing the session to be reconnected
it is possible we will leak -EAGAIN back to the application even for
system calls such as stat() where this is not a valid error.

Fix this by re-trying the operation from within cifs_revalidate_dentry_attr()
if cifs_get_inode_info*() returns -EAGAIN.

This fixes stat() and possibly also other system calls that uses
cifs_revalidate_dentry*().

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
---
 fs/cifs/inode.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index b5e6635c578e..1c6f659110d0 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -2073,6 +2073,7 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
 	struct inode *inode = d_inode(dentry);
 	struct super_block *sb = dentry->d_sb;
 	char *full_path = NULL;
+	int count = 0;
 
 	if (inode == NULL)
 		return -ENOENT;
@@ -2094,15 +2095,18 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
 		 full_path, inode, inode->i_count.counter,
 		 dentry, cifs_get_time(dentry), jiffies);
 
+again:
 	if (cifs_sb_master_tcon(CIFS_SB(sb))->unix_ext)
 		rc = cifs_get_inode_info_unix(&inode, full_path, sb, xid);
 	else
 		rc = cifs_get_inode_info(&inode, full_path, NULL, sb,
 					 xid, NULL);
-
+	if (rc == -EAGAIN && count++ < 10)
+		goto again;
 out:
 	kfree(full_path);
 	free_xid(xid);
+
 	return rc;
 }
 
-- 
2.13.6

Re: [PATCH] cifs: don't leak -EAGAIN for stat() during reconnect

[Aurélien Aptel]

Hi Ronnie,

Ronnie Sahlberg <lsahlber@redhat.com> writes:
> Fix this by re-trying the operation from within cifs_revalidate_dentry_attr()
> if cifs_get_inode_info*() returns -EAGAIN.

Would it make sense to use is_retryable_error() instead of checking for
just EAGAIN? It also checks for interruption errors.

Cheers,
-- 
Aurélien Aptel / SUSE Labs Samba Team
GPG: 1839 CB5F 9F5B FB9B AA97  8C99 03C8 A49B 521B D5D3
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg, DE
GF: Felix Imendörffer, Mary Higgins, Sri Rasiah HRB 247165 (AG München)

Re: [PATCH] cifs: don't leak -EAGAIN for stat() during reconnect

[ronnie sahlberg]

On Tue, Feb 18, 2020 at 9:47 PM Aurélien Aptel <aaptel@suse.com> wrote:
>
> Hi Ronnie,
>
> Ronnie Sahlberg <lsahlber@redhat.com> writes:
> > Fix this by re-trying the operation from within cifs_revalidate_dentry_attr()
> > if cifs_get_inode_info*() returns -EAGAIN.
>
> Would it make sense to use is_retryable_error() instead of checking for
> just EAGAIN? It also checks for interruption errors.

Makes sense. Thanks.

>
> Cheers,
> --
> Aurélien Aptel / SUSE Labs Samba Team
> GPG: 1839 CB5F 9F5B FB9B AA97  8C99 03C8 A49B 521B D5D3
> SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg, DE
> GF: Felix Imendörffer, Mary Higgins, Sri Rasiah HRB 247165 (AG München)

[PATCH] cifs: don't leak -EAGAIN for stat() during reconnect

[Ronnie Sahlberg]

If from cifs_revalidate_dentry_attr() the SMB2/QUERY_INFO call fails with an
error, such as STATUS_SESSION_EXPIRED, causing the session to be reconnected
it is possible we will leak -EAGAIN back to the application even for
system calls such as stat() where this is not a valid error.

Fix this by re-trying the operation from within cifs_revalidate_dentry_attr()
if cifs_get_inode_info*() returns -EAGAIN.

This fixes stat() and possibly also other system calls that uses
cifs_revalidate_dentry*().

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
---
 fs/cifs/inode.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index b5e6635c578e..1212ace05258 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -2073,6 +2073,7 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
 	struct inode *inode = d_inode(dentry);
 	struct super_block *sb = dentry->d_sb;
 	char *full_path = NULL;
+	int count = 0;
 
 	if (inode == NULL)
 		return -ENOENT;
@@ -2094,15 +2095,18 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
 		 full_path, inode, inode->i_count.counter,
 		 dentry, cifs_get_time(dentry), jiffies);
 
+again:
 	if (cifs_sb_master_tcon(CIFS_SB(sb))->unix_ext)
 		rc = cifs_get_inode_info_unix(&inode, full_path, sb, xid);
 	else
 		rc = cifs_get_inode_info(&inode, full_path, NULL, sb,
 					 xid, NULL);
-
+	if (is_retryable_error(rc) && count++ < 10)
+		goto again;
 out:
 	kfree(full_path);
 	free_xid(xid);
+
 	return rc;
 }
 
-- 
2.13.6

Re: [PATCH] cifs: don't leak -EAGAIN for stat() during reconnect

[Steve French]

merged into cifs-2.6.git for-next

On Tue, Feb 18, 2020 at 2:07 PM Ronnie Sahlberg <lsahlber@redhat.com> wrote:
>
> If from cifs_revalidate_dentry_attr() the SMB2/QUERY_INFO call fails with an
> error, such as STATUS_SESSION_EXPIRED, causing the session to be reconnected
> it is possible we will leak -EAGAIN back to the application even for
> system calls such as stat() where this is not a valid error.
>
> Fix this by re-trying the operation from within cifs_revalidate_dentry_attr()
> if cifs_get_inode_info*() returns -EAGAIN.
>
> This fixes stat() and possibly also other system calls that uses
> cifs_revalidate_dentry*().
>
> Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
> ---
>  fs/cifs/inode.c | 6 +++++-
>  1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
> index b5e6635c578e..1212ace05258 100644
> --- a/fs/cifs/inode.c
> +++ b/fs/cifs/inode.c
> @@ -2073,6 +2073,7 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
>         struct inode *inode = d_inode(dentry);
>         struct super_block *sb = dentry->d_sb;
>         char *full_path = NULL;
> +       int count = 0;
>
>         if (inode == NULL)
>                 return -ENOENT;
> @@ -2094,15 +2095,18 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
>                  full_path, inode, inode->i_count.counter,
>                  dentry, cifs_get_time(dentry), jiffies);
>
> +again:
>         if (cifs_sb_master_tcon(CIFS_SB(sb))->unix_ext)
>                 rc = cifs_get_inode_info_unix(&inode, full_path, sb, xid);
>         else
>                 rc = cifs_get_inode_info(&inode, full_path, NULL, sb,
>                                          xid, NULL);
> -
> +       if (is_retryable_error(rc) && count++ < 10)
> +               goto again;
>  out:
>         kfree(full_path);
>         free_xid(xid);
> +
>         return rc;
>  }
>
> --
> 2.13.6
>


-- 
Thanks,

Steve

Re: [PATCH] cifs: don't leak -EAGAIN for stat() during reconnect

[Pavel Shilovsky]

вт, 18 февр. 2020 г. в 15:27, Steve French <smfrench@gmail.com>:
>
> merged into cifs-2.6.git for-next
>
> On Tue, Feb 18, 2020 at 2:07 PM Ronnie Sahlberg <lsahlber@redhat.com> wrote:
> >
> > If from cifs_revalidate_dentry_attr() the SMB2/QUERY_INFO call fails with an
> > error, such as STATUS_SESSION_EXPIRED, causing the session to be reconnected
> > it is possible we will leak -EAGAIN back to the application even for
> > system calls such as stat() where this is not a valid error.
> >
> > Fix this by re-trying the operation from within cifs_revalidate_dentry_attr()
> > if cifs_get_inode_info*() returns -EAGAIN.
> >
> > This fixes stat() and possibly also other system calls that uses
> > cifs_revalidate_dentry*().
> >
> > Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
> > ---
> >  fs/cifs/inode.c | 6 +++++-
> >  1 file changed, 5 insertions(+), 1 deletion(-)
> >
> > diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
> > index b5e6635c578e..1212ace05258 100644
> > --- a/fs/cifs/inode.c
> > +++ b/fs/cifs/inode.c
> > @@ -2073,6 +2073,7 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
> >         struct inode *inode = d_inode(dentry);
> >         struct super_block *sb = dentry->d_sb;
> >         char *full_path = NULL;
> > +       int count = 0;
> >
> >         if (inode == NULL)
> >                 return -ENOENT;
> > @@ -2094,15 +2095,18 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
> >                  full_path, inode, inode->i_count.counter,
> >                  dentry, cifs_get_time(dentry), jiffies);
> >
> > +again:
> >         if (cifs_sb_master_tcon(CIFS_SB(sb))->unix_ext)
> >                 rc = cifs_get_inode_info_unix(&inode, full_path, sb, xid);
> >         else
> >                 rc = cifs_get_inode_info(&inode, full_path, NULL, sb,
> >                                          xid, NULL);
> > -
> > +       if (is_retryable_error(rc) && count++ < 10)
> > +               goto again;

If there is interrupt error, you will end up doing 10 attempts with
the same outcome - interrupt error. Such errors should be returned to
the upper layers to be handled correctly (restart of a system call or
return of EINTR error to the user space).

Please revert to your original version that handles EAGAIN only.

--
Best regards,
Pavel Shilovsky

> >  out:
> >         kfree(full_path);
> >         free_xid(xid);
> > +
> >         return rc;
> >  }
> >
> > --
> > 2.13.6
> >
>
>
> --
> Thanks,
>
> Steve

Re: [PATCH] cifs: don't leak -EAGAIN for stat() during reconnect

[Steve French]

[-- Attachment #1: Type: text/plain, Size: 2897 bytes --]

updated Ronnie's patch and remerged to cifs-2.6.git for-next

Let me know if any objections

On Mon, Feb 24, 2020 at 1:36 PM Pavel Shilovsky <piastryyy@gmail.com> wrote:
>
> вт, 18 февр. 2020 г. в 15:27, Steve French <smfrench@gmail.com>:
> >
> > merged into cifs-2.6.git for-next
> >
> > On Tue, Feb 18, 2020 at 2:07 PM Ronnie Sahlberg <lsahlber@redhat.com> wrote:
> > >
> > > If from cifs_revalidate_dentry_attr() the SMB2/QUERY_INFO call fails with an
> > > error, such as STATUS_SESSION_EXPIRED, causing the session to be reconnected
> > > it is possible we will leak -EAGAIN back to the application even for
> > > system calls such as stat() where this is not a valid error.
> > >
> > > Fix this by re-trying the operation from within cifs_revalidate_dentry_attr()
> > > if cifs_get_inode_info*() returns -EAGAIN.
> > >
> > > This fixes stat() and possibly also other system calls that uses
> > > cifs_revalidate_dentry*().
> > >
> > > Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
> > > ---
> > >  fs/cifs/inode.c | 6 +++++-
> > >  1 file changed, 5 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
> > > index b5e6635c578e..1212ace05258 100644
> > > --- a/fs/cifs/inode.c
> > > +++ b/fs/cifs/inode.c
> > > @@ -2073,6 +2073,7 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
> > >         struct inode *inode = d_inode(dentry);
> > >         struct super_block *sb = dentry->d_sb;
> > >         char *full_path = NULL;
> > > +       int count = 0;
> > >
> > >         if (inode == NULL)
> > >                 return -ENOENT;
> > > @@ -2094,15 +2095,18 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
> > >                  full_path, inode, inode->i_count.counter,
> > >                  dentry, cifs_get_time(dentry), jiffies);
> > >
> > > +again:
> > >         if (cifs_sb_master_tcon(CIFS_SB(sb))->unix_ext)
> > >                 rc = cifs_get_inode_info_unix(&inode, full_path, sb, xid);
> > >         else
> > >                 rc = cifs_get_inode_info(&inode, full_path, NULL, sb,
> > >                                          xid, NULL);
> > > -
> > > +       if (is_retryable_error(rc) && count++ < 10)
> > > +               goto again;
>
> If there is interrupt error, you will end up doing 10 attempts with
> the same outcome - interrupt error. Such errors should be returned to
> the upper layers to be handled correctly (restart of a system call or
> return of EINTR error to the user space).
>
> Please revert to your original version that handles EAGAIN only.
>
> --
> Best regards,
> Pavel Shilovsky
>
> > >  out:
> > >         kfree(full_path);
> > >         free_xid(xid);
> > > +
> > >         return rc;
> > >  }
> > >
> > > --
> > > 2.13.6
> > >
> >
> >
> > --
> > Thanks,
> >
> > Steve



-- 
Thanks,

Steve

[-- Attachment #2: 0001-cifs-don-t-leak-EAGAIN-for-stat-during-reconnect.patch --]
[-- Type: text/x-patch, Size: 1931 bytes --]

From 9255f1c5788c0d97d0ef31b0fabb6457787ce680 Mon Sep 17 00:00:00 2001
From: Ronnie Sahlberg <lsahlber@redhat.com>
Date: Wed, 19 Feb 2020 06:01:03 +1000
Subject: [PATCH 01/10] cifs: don't leak -EAGAIN for stat() during reconnect

If from cifs_revalidate_dentry_attr() the SMB2/QUERY_INFO call fails with an
error, such as STATUS_SESSION_EXPIRED, causing the session to be reconnected
it is possible we will leak -EAGAIN back to the application even for
system calls such as stat() where this is not a valid error.

Fix this by re-trying the operation from within cifs_revalidate_dentry_attr()
if cifs_get_inode_info*() returns -EAGAIN.

This fixes stat() and possibly also other system calls that uses
cifs_revalidate_dentry*().

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
CC: Stable <stable@vger.kernel.org>
---
 fs/cifs/inode.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index b5e6635c578e..1212ace05258 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -2073,6 +2073,7 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
 	struct inode *inode = d_inode(dentry);
 	struct super_block *sb = dentry->d_sb;
 	char *full_path = NULL;
+	int count = 0;
 
 	if (inode == NULL)
 		return -ENOENT;
@@ -2094,15 +2095,18 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
 		 full_path, inode, inode->i_count.counter,
 		 dentry, cifs_get_time(dentry), jiffies);
 
+again:
 	if (cifs_sb_master_tcon(CIFS_SB(sb))->unix_ext)
 		rc = cifs_get_inode_info_unix(&inode, full_path, sb, xid);
 	else
 		rc = cifs_get_inode_info(&inode, full_path, NULL, sb,
 					 xid, NULL);
-
+	if (rc == -EAGAIN && count++ < 10)
+		goto again;
 out:
 	kfree(full_path);
 	free_xid(xid);
+
 	return rc;
 }
 
-- 
2.20.1

[PATCH 0/1] dont leak -EAGAIN

[Ronnie Sahlberg]

Steve, List

Please find a small patch to fix an issue where during a session reconnect we may leak
-EAGAIN back to the application instead of retrying the operation once reconnect has completed.

This can affects for example the stat() call :
  stat("/mnt", 0x55b802c096f8) = -1 EAGAIN (Resource temporarily unavailable) <0.002447>
causing applications such as 'ls' to fail with an error as this is not a valid return
for stat()

[PATCH] cifs: don't leak -EAGAIN for stat() during reconnect

[Ronnie Sahlberg]

If from cifs_revalidate_dentry_attr() the SMB2/QUERY_INFO call fails with an
error, such as STATUS_SESSION_EXPIRED, causing the session to be reconnected
it is possible we will leak -EAGAIN back to the application even for
system calls such as stat() where this is not a valid error.

Fix this by re-trying the operation from within cifs_revalidate_dentry_attr()
if cifs_get_inode_info*() returns -EAGAIN.

This fixes stat() and possibly also other system calls that uses
cifs_revalidate_dentry*().

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
---
 fs/cifs/inode.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/fs/cifs/inode.c b/fs/cifs/inode.c
index b5e6635c578e..cf36dcd9dafd 100644
--- a/fs/cifs/inode.c
+++ b/fs/cifs/inode.c
@@ -2073,7 +2073,9 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
 	struct inode *inode = d_inode(dentry);
 	struct super_block *sb = dentry->d_sb;
 	char *full_path = NULL;
+	int count = 0;
 
+again:
 	if (inode == NULL)
 		return -ENOENT;
 
@@ -2099,6 +2101,8 @@ int cifs_revalidate_dentry_attr(struct dentry *dentry)
 	else
 		rc = cifs_get_inode_info(&inode, full_path, NULL, sb,
 					 xid, NULL);
+	if (rc == -EAGAIN && count++ < 10)
+		goto again;
 
 out:
 	kfree(full_path);
-- 
2.13.6