ceph: fix up endian bug in managing feature bits
diff mbox series

Message ID 1588023986-23672-1-git-send-email-edward6@linux.ibm.com
State New
Headers show
Series
  • ceph: fix up endian bug in managing feature bits
Related show

Commit Message

Eduard Shishkin April 27, 2020, 9:46 p.m. UTC
From: Eduard Shishkin <edward6@linux.ibm.com>

In the function handle_session() variable @features always
contains little endian order of bytes. Just because the feature
bits are packed bytewise from left to right in
encode_supported_features().

However, test_bit(), called to check features availability, assumes
the host order of bytes in that variable. This leads to problems on
big endian architectures. Specifically it is impossible to mount
ceph volume on s390.

This patch adds conversion from little endian to the host order
of bytes, thus fixing the problem.

Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
---
 fs/ceph/mds_client.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Jeff Layton April 28, 2020, 12:23 p.m. UTC | #1
On Mon, 2020-04-27 at 23:46 +0200, edward6@linux.ibm.com wrote:
> From: Eduard Shishkin <edward6@linux.ibm.com>
> 
> In the function handle_session() variable @features always
> contains little endian order of bytes. Just because the feature
> bits are packed bytewise from left to right in
> encode_supported_features().
> 
> However, test_bit(), called to check features availability, assumes
> the host order of bytes in that variable. This leads to problems on
> big endian architectures. Specifically it is impossible to mount
> ceph volume on s390.
> 
> This patch adds conversion from little endian to the host order
> of bytes, thus fixing the problem.
> 
> Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
> ---
>  fs/ceph/mds_client.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> index 486f91f..190598d 100644
> --- a/fs/ceph/mds_client.c
> +++ b/fs/ceph/mds_client.c
> @@ -3252,7 +3252,7 @@ static void handle_session(struct ceph_mds_session *session,
>  	struct ceph_mds_session_head *h;
>  	u32 op;
>  	u64 seq;
> -	unsigned long features = 0;
> +	__le64 features = 0;
>  	int wake = 0;
>  	bool blacklisted = false;
>  
> @@ -3301,7 +3301,7 @@ static void handle_session(struct ceph_mds_session *session,
>  		if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
>  			pr_info("mds%d reconnect success\n", session->s_mds);
>  		session->s_state = CEPH_MDS_SESSION_OPEN;
> -		session->s_features = features;
> +		session->s_features = le64_to_cpu(features);
>  		renewed_caps(mdsc, session, 0);
>  		wake = 1;
>  		if (mdsc->stopping)

(cc'ing Zheng since he did the original patches here)

Thanks Eduard. The problem is real, but I think we can just do the
conversion during the decode.

The feature mask words sent by the MDS are 64 bits, so if it's smaller
we can assume that it's malformed. So, I don't think we need to handle
the case where it's smaller than 8 bytes.

How about this patch instead?

--------------------------8<-----------------------------

ceph: fix endianness bug when handling MDS session feature bits

Eduard reported a problem mounting cephfs on s390 arch. The feature
mask sent by the MDS is little-endian, so we need to convert it
before storing and testing against it.

Reported-by: Eduard Shishkin <edward6@linux.ibm.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
---
 fs/ceph/mds_client.c | 8 +++-----
 1 file changed, 3 insertions(+), 5 deletions(-)

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index a8a5b98148ec..6c283c52d401 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -3260,8 +3260,7 @@ static void handle_session(struct ceph_mds_session *session,
 	void *end = p + msg->front.iov_len;
 	struct ceph_mds_session_head *h;
 	u32 op;
-	u64 seq;
-	unsigned long features = 0;
+	u64 seq, features = 0;
 	int wake = 0;
 	bool blacklisted = false;
 
@@ -3280,9 +3279,8 @@ static void handle_session(struct ceph_mds_session *session,
 			goto bad;
 		/* version >= 3, feature bits */
 		ceph_decode_32_safe(&p, end, len, bad);
-		ceph_decode_need(&p, end, len, bad);
-		memcpy(&features, p, min_t(size_t, len, sizeof(features)));
-		p += len;
+		ceph_decode_64_safe(&p, end, features, bad);
+		p += len - sizeof(features);
 	}
 
 	mutex_lock(&mdsc->mutex);
Yan, Zheng April 28, 2020, 12:44 p.m. UTC | #2
On Tue, Apr 28, 2020 at 8:23 PM Jeff Layton <jlayton@kernel.org> wrote:
>
> On Mon, 2020-04-27 at 23:46 +0200, edward6@linux.ibm.com wrote:
> > From: Eduard Shishkin <edward6@linux.ibm.com>
> >
> > In the function handle_session() variable @features always
> > contains little endian order of bytes. Just because the feature
> > bits are packed bytewise from left to right in
> > encode_supported_features().
> >
> > However, test_bit(), called to check features availability, assumes
> > the host order of bytes in that variable. This leads to problems on
> > big endian architectures. Specifically it is impossible to mount
> > ceph volume on s390.
> >
> > This patch adds conversion from little endian to the host order
> > of bytes, thus fixing the problem.
> >
> > Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
> > ---
> >  fs/ceph/mds_client.c | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> > index 486f91f..190598d 100644
> > --- a/fs/ceph/mds_client.c
> > +++ b/fs/ceph/mds_client.c
> > @@ -3252,7 +3252,7 @@ static void handle_session(struct ceph_mds_session *session,
> >       struct ceph_mds_session_head *h;
> >       u32 op;
> >       u64 seq;
> > -     unsigned long features = 0;
> > +     __le64 features = 0;
> >       int wake = 0;
> >       bool blacklisted = false;
> >
> > @@ -3301,7 +3301,7 @@ static void handle_session(struct ceph_mds_session *session,
> >               if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
> >                       pr_info("mds%d reconnect success\n", session->s_mds);
> >               session->s_state = CEPH_MDS_SESSION_OPEN;
> > -             session->s_features = features;
> > +             session->s_features = le64_to_cpu(features);
> >               renewed_caps(mdsc, session, 0);
> >               wake = 1;
> >               if (mdsc->stopping)
>
> (cc'ing Zheng since he did the original patches here)
>
> Thanks Eduard. The problem is real, but I think we can just do the
> conversion during the decode.
>
> The feature mask words sent by the MDS are 64 bits, so if it's smaller
> we can assume that it's malformed. So, I don't think we need to handle
> the case where it's smaller than 8 bytes.
>
> How about this patch instead?
>
> --------------------------8<-----------------------------
>
> ceph: fix endianness bug when handling MDS session feature bits
>
> Eduard reported a problem mounting cephfs on s390 arch. The feature
> mask sent by the MDS is little-endian, so we need to convert it
> before storing and testing against it.
>
> Reported-by: Eduard Shishkin <edward6@linux.ibm.com>
> Signed-off-by: Jeff Layton <jlayton@kernel.org>
> ---
>  fs/ceph/mds_client.c | 8 +++-----
>  1 file changed, 3 insertions(+), 5 deletions(-)
>
> diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> index a8a5b98148ec..6c283c52d401 100644
> --- a/fs/ceph/mds_client.c
> +++ b/fs/ceph/mds_client.c
> @@ -3260,8 +3260,7 @@ static void handle_session(struct ceph_mds_session *session,
>         void *end = p + msg->front.iov_len;
>         struct ceph_mds_session_head *h;
>         u32 op;
> -       u64 seq;
> -       unsigned long features = 0;
> +       u64 seq, features = 0;
>         int wake = 0;
>         bool blacklisted = false;
>
> @@ -3280,9 +3279,8 @@ static void handle_session(struct ceph_mds_session *session,
>                         goto bad;
>                 /* version >= 3, feature bits */
>                 ceph_decode_32_safe(&p, end, len, bad);
> -               ceph_decode_need(&p, end, len, bad);
> -               memcpy(&features, p, min_t(size_t, len, sizeof(features)));
> -               p += len;
> +               ceph_decode_64_safe(&p, end, features, bad);
> +               p += len - sizeof(features);
>         }

this one looks better.

Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
>
>         mutex_lock(&mdsc->mutex);
> --
> 2.26.1
>
>
Eduard Shishkin April 29, 2020, 9:46 a.m. UTC | #3
On 4/28/20 2:23 PM, Jeff Layton wrote:
> On Mon, 2020-04-27 at 23:46 +0200, edward6@linux.ibm.com wrote:
>> From: Eduard Shishkin <edward6@linux.ibm.com>
>>
>> In the function handle_session() variable @features always
>> contains little endian order of bytes. Just because the feature
>> bits are packed bytewise from left to right in
>> encode_supported_features().
>>
>> However, test_bit(), called to check features availability, assumes
>> the host order of bytes in that variable. This leads to problems on
>> big endian architectures. Specifically it is impossible to mount
>> ceph volume on s390.
>>
>> This patch adds conversion from little endian to the host order
>> of bytes, thus fixing the problem.
>>
>> Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
>> ---
>>   fs/ceph/mds_client.c | 4 ++--
>>   1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
>> index 486f91f..190598d 100644
>> --- a/fs/ceph/mds_client.c
>> +++ b/fs/ceph/mds_client.c
>> @@ -3252,7 +3252,7 @@ static void handle_session(struct ceph_mds_session *session,
>>   	struct ceph_mds_session_head *h;
>>   	u32 op;
>>   	u64 seq;
>> -	unsigned long features = 0;
>> +	__le64 features = 0;
>>   	int wake = 0;
>>   	bool blacklisted = false;
>>   
>> @@ -3301,7 +3301,7 @@ static void handle_session(struct ceph_mds_session *session,
>>   		if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
>>   			pr_info("mds%d reconnect success\n", session->s_mds);
>>   		session->s_state = CEPH_MDS_SESSION_OPEN;
>> -		session->s_features = features;
>> +		session->s_features = le64_to_cpu(features);
>>   		renewed_caps(mdsc, session, 0);
>>   		wake = 1;
>>   		if (mdsc->stopping)
> 
> (cc'ing Zheng since he did the original patches here)
> 
> Thanks Eduard. The problem is real, but I think we can just do the
> conversion during the decode.
> 
> The feature mask words sent by the MDS are 64 bits, so if it's smaller
> we can assume that it's malformed. So, I don't think we need to handle
> the case where it's smaller than 8 bytes.
> 
> How about this patch instead?


Hi Jeff,

This also works. Please, apply.

Thanks,
Eduard.

> 
> --------------------------8<-----------------------------
> 
> ceph: fix endianness bug when handling MDS session feature bits
> 
> Eduard reported a problem mounting cephfs on s390 arch. The feature
> mask sent by the MDS is little-endian, so we need to convert it
> before storing and testing against it.
> 
> Reported-by: Eduard Shishkin <edward6@linux.ibm.com>
> Signed-off-by: Jeff Layton <jlayton@kernel.org>
> ---
>   fs/ceph/mds_client.c | 8 +++-----
>   1 file changed, 3 insertions(+), 5 deletions(-)
> 
> diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> index a8a5b98148ec..6c283c52d401 100644
> --- a/fs/ceph/mds_client.c
> +++ b/fs/ceph/mds_client.c
> @@ -3260,8 +3260,7 @@ static void handle_session(struct ceph_mds_session *session,
>   	void *end = p + msg->front.iov_len;
>   	struct ceph_mds_session_head *h;
>   	u32 op;
> -	u64 seq;
> -	unsigned long features = 0;
> +	u64 seq, features = 0;
>   	int wake = 0;
>   	bool blacklisted = false;
>   
> @@ -3280,9 +3279,8 @@ static void handle_session(struct ceph_mds_session *session,
>   			goto bad;
>   		/* version >= 3, feature bits */
>   		ceph_decode_32_safe(&p, end, len, bad);
> -		ceph_decode_need(&p, end, len, bad);
> -		memcpy(&features, p, min_t(size_t, len, sizeof(features)));
> -		p += len;
> +		ceph_decode_64_safe(&p, end, features, bad);
> +		p += len - sizeof(features);
>   	}
>   
>   	mutex_lock(&mdsc->mutex);
>
Jeff Layton April 29, 2020, 3:39 p.m. UTC | #4
On Wed, 2020-04-29 at 11:46 +0200, Eduard Shishkin wrote:
> On 4/28/20 2:23 PM, Jeff Layton wrote:
> > On Mon, 2020-04-27 at 23:46 +0200, edward6@linux.ibm.com wrote:
> > > From: Eduard Shishkin <edward6@linux.ibm.com>
> > > 
> > > In the function handle_session() variable @features always
> > > contains little endian order of bytes. Just because the feature
> > > bits are packed bytewise from left to right in
> > > encode_supported_features().
> > > 
> > > However, test_bit(), called to check features availability, assumes
> > > the host order of bytes in that variable. This leads to problems on
> > > big endian architectures. Specifically it is impossible to mount
> > > ceph volume on s390.
> > > 
> > > This patch adds conversion from little endian to the host order
> > > of bytes, thus fixing the problem.
> > > 
> > > Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
> > > ---
> > >   fs/ceph/mds_client.c | 4 ++--
> > >   1 file changed, 2 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> > > index 486f91f..190598d 100644
> > > --- a/fs/ceph/mds_client.c
> > > +++ b/fs/ceph/mds_client.c
> > > @@ -3252,7 +3252,7 @@ static void handle_session(struct ceph_mds_session *session,
> > >   	struct ceph_mds_session_head *h;
> > >   	u32 op;
> > >   	u64 seq;
> > > -	unsigned long features = 0;
> > > +	__le64 features = 0;
> > >   	int wake = 0;
> > >   	bool blacklisted = false;
> > >   
> > > @@ -3301,7 +3301,7 @@ static void handle_session(struct ceph_mds_session *session,
> > >   		if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
> > >   			pr_info("mds%d reconnect success\n", session->s_mds);
> > >   		session->s_state = CEPH_MDS_SESSION_OPEN;
> > > -		session->s_features = features;
> > > +		session->s_features = le64_to_cpu(features);
> > >   		renewed_caps(mdsc, session, 0);
> > >   		wake = 1;
> > >   		if (mdsc->stopping)
> > 
> > (cc'ing Zheng since he did the original patches here)
> > 
> > Thanks Eduard. The problem is real, but I think we can just do the
> > conversion during the decode.
> > 
> > The feature mask words sent by the MDS are 64 bits, so if it's smaller
> > we can assume that it's malformed. So, I don't think we need to handle
> > the case where it's smaller than 8 bytes.
> > 
> > How about this patch instead?
> 
> Hi Jeff,
> 
> This also works. Please, apply.
> 
> Thanks,
> Eduard.
> 

Thanks. Merged into ceph-client/testing branch, and should make v5.8.

> > --------------------------8<-----------------------------
> > 
> > ceph: fix endianness bug when handling MDS session feature bits
> > 
> > Eduard reported a problem mounting cephfs on s390 arch. The feature
> > mask sent by the MDS is little-endian, so we need to convert it
> > before storing and testing against it.
> > 
> > Reported-by: Eduard Shishkin <edward6@linux.ibm.com>
> > Signed-off-by: Jeff Layton <jlayton@kernel.org>
> > ---
> >   fs/ceph/mds_client.c | 8 +++-----
> >   1 file changed, 3 insertions(+), 5 deletions(-)
> > 
> > diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> > index a8a5b98148ec..6c283c52d401 100644
> > --- a/fs/ceph/mds_client.c
> > +++ b/fs/ceph/mds_client.c
> > @@ -3260,8 +3260,7 @@ static void handle_session(struct ceph_mds_session *session,
> >   	void *end = p + msg->front.iov_len;
> >   	struct ceph_mds_session_head *h;
> >   	u32 op;
> > -	u64 seq;
> > -	unsigned long features = 0;
> > +	u64 seq, features = 0;
> >   	int wake = 0;
> >   	bool blacklisted = false;
> >   
> > @@ -3280,9 +3279,8 @@ static void handle_session(struct ceph_mds_session *session,
> >   			goto bad;
> >   		/* version >= 3, feature bits */
> >   		ceph_decode_32_safe(&p, end, len, bad);
> > -		ceph_decode_need(&p, end, len, bad);
> > -		memcpy(&features, p, min_t(size_t, len, sizeof(features)));
> > -		p += len;
> > +		ceph_decode_64_safe(&p, end, features, bad);
> > +		p += len - sizeof(features);
> >   	}
> >   
> >   	mutex_lock(&mdsc->mutex);
> >
Ilya Dryomov April 29, 2020, 4:08 p.m. UTC | #5
On Wed, Apr 29, 2020 at 5:42 PM Jeff Layton <jlayton@kernel.org> wrote:
>
> On Wed, 2020-04-29 at 11:46 +0200, Eduard Shishkin wrote:
> > On 4/28/20 2:23 PM, Jeff Layton wrote:
> > > On Mon, 2020-04-27 at 23:46 +0200, edward6@linux.ibm.com wrote:
> > > > From: Eduard Shishkin <edward6@linux.ibm.com>
> > > >
> > > > In the function handle_session() variable @features always
> > > > contains little endian order of bytes. Just because the feature
> > > > bits are packed bytewise from left to right in
> > > > encode_supported_features().
> > > >
> > > > However, test_bit(), called to check features availability, assumes
> > > > the host order of bytes in that variable. This leads to problems on
> > > > big endian architectures. Specifically it is impossible to mount
> > > > ceph volume on s390.
> > > >
> > > > This patch adds conversion from little endian to the host order
> > > > of bytes, thus fixing the problem.
> > > >
> > > > Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
> > > > ---
> > > >   fs/ceph/mds_client.c | 4 ++--
> > > >   1 file changed, 2 insertions(+), 2 deletions(-)
> > > >
> > > > diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> > > > index 486f91f..190598d 100644
> > > > --- a/fs/ceph/mds_client.c
> > > > +++ b/fs/ceph/mds_client.c
> > > > @@ -3252,7 +3252,7 @@ static void handle_session(struct ceph_mds_session *session,
> > > >           struct ceph_mds_session_head *h;
> > > >           u32 op;
> > > >           u64 seq;
> > > > - unsigned long features = 0;
> > > > + __le64 features = 0;
> > > >           int wake = 0;
> > > >           bool blacklisted = false;
> > > >
> > > > @@ -3301,7 +3301,7 @@ static void handle_session(struct ceph_mds_session *session,
> > > >                   if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
> > > >                           pr_info("mds%d reconnect success\n", session->s_mds);
> > > >                   session->s_state = CEPH_MDS_SESSION_OPEN;
> > > > -         session->s_features = features;
> > > > +         session->s_features = le64_to_cpu(features);
> > > >                   renewed_caps(mdsc, session, 0);
> > > >                   wake = 1;
> > > >                   if (mdsc->stopping)
> > >
> > > (cc'ing Zheng since he did the original patches here)
> > >
> > > Thanks Eduard. The problem is real, but I think we can just do the
> > > conversion during the decode.
> > >
> > > The feature mask words sent by the MDS are 64 bits, so if it's smaller
> > > we can assume that it's malformed. So, I don't think we need to handle
> > > the case where it's smaller than 8 bytes.
> > >
> > > How about this patch instead?
> >
> > Hi Jeff,
> >
> > This also works. Please, apply.
> >
> > Thanks,
> > Eduard.
> >
>
> Thanks. Merged into ceph-client/testing branch, and should make v5.8.

I think this is stable material.  I'll tag it and get it queued up for 5.7-rc.

Thanks,

                Ilya
Jeff Layton April 30, 2020, 1:03 p.m. UTC | #6
On Wed, 2020-04-29 at 18:08 +0200, Ilya Dryomov wrote:
> On Wed, Apr 29, 2020 at 5:42 PM Jeff Layton <jlayton@kernel.org> wrote:
> > On Wed, 2020-04-29 at 11:46 +0200, Eduard Shishkin wrote:
> > > On 4/28/20 2:23 PM, Jeff Layton wrote:
> > > > On Mon, 2020-04-27 at 23:46 +0200, edward6@linux.ibm.com wrote:
> > > > > From: Eduard Shishkin <edward6@linux.ibm.com>
> > > > > 
> > > > > In the function handle_session() variable @features always
> > > > > contains little endian order of bytes. Just because the feature
> > > > > bits are packed bytewise from left to right in
> > > > > encode_supported_features().
> > > > > 
> > > > > However, test_bit(), called to check features availability, assumes
> > > > > the host order of bytes in that variable. This leads to problems on
> > > > > big endian architectures. Specifically it is impossible to mount
> > > > > ceph volume on s390.
> > > > > 
> > > > > This patch adds conversion from little endian to the host order
> > > > > of bytes, thus fixing the problem.
> > > > > 
> > > > > Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
> > > > > ---
> > > > >   fs/ceph/mds_client.c | 4 ++--
> > > > >   1 file changed, 2 insertions(+), 2 deletions(-)
> > > > > 
> > > > > diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> > > > > index 486f91f..190598d 100644
> > > > > --- a/fs/ceph/mds_client.c
> > > > > +++ b/fs/ceph/mds_client.c
> > > > > @@ -3252,7 +3252,7 @@ static void handle_session(struct ceph_mds_session *session,
> > > > >           struct ceph_mds_session_head *h;
> > > > >           u32 op;
> > > > >           u64 seq;
> > > > > - unsigned long features = 0;
> > > > > + __le64 features = 0;
> > > > >           int wake = 0;
> > > > >           bool blacklisted = false;
> > > > > 
> > > > > @@ -3301,7 +3301,7 @@ static void handle_session(struct ceph_mds_session *session,
> > > > >                   if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
> > > > >                           pr_info("mds%d reconnect success\n", session->s_mds);
> > > > >                   session->s_state = CEPH_MDS_SESSION_OPEN;
> > > > > -         session->s_features = features;
> > > > > +         session->s_features = le64_to_cpu(features);
> > > > >                   renewed_caps(mdsc, session, 0);
> > > > >                   wake = 1;
> > > > >                   if (mdsc->stopping)
> > > > 
> > > > (cc'ing Zheng since he did the original patches here)
> > > > 
> > > > Thanks Eduard. The problem is real, but I think we can just do the
> > > > conversion during the decode.
> > > > 
> > > > The feature mask words sent by the MDS are 64 bits, so if it's smaller
> > > > we can assume that it's malformed. So, I don't think we need to handle
> > > > the case where it's smaller than 8 bytes.
> > > > 
> > > > How about this patch instead?
> > > 
> > > Hi Jeff,
> > > 
> > > This also works. Please, apply.
> > > 
> > > Thanks,
> > > Eduard.
> > > 
> > 
> > Thanks. Merged into ceph-client/testing branch, and should make v5.8.
> 
> I think this is stable material.  I'll tag it and get it queued up for 5.7-rc.
> 
> Thanks,
> 

Yeah, that sounds reasonable.

If you're going to send up another PR, then we might want to add these
bugfixes currently in the testing branch to it as well:

445645c8be5f fs/ceph:fix special error code in ceph_try_get_caps()
591681748b56 fs/ceph:fix double unlock in handle_cap_export()
0e84a1ebe161 ceph: ceph_kick_flushing_caps needs the s_mutex

I'm not sure that any of them need to go to stable though. We might also
want this one though:

7b3facb61440 ceph: reset i_requested_max_size if file write is not wanted

...but it'll probably need to be reworked due to merge conflicts if we
move it ahead of some of the cap handling cleanup patches (or we could
just pull those in too).

Zheng, do you have an opinion here? Should 7b3facb61440 go to stable?

Thanks,
--
Jeff Layton <jlayton@kernel.org>
Ilya Dryomov May 4, 2020, 5:48 p.m. UTC | #7
On Thu, Apr 30, 2020 at 3:03 PM Jeff Layton <jlayton@kernel.org> wrote:
>
> On Wed, 2020-04-29 at 18:08 +0200, Ilya Dryomov wrote:
> > On Wed, Apr 29, 2020 at 5:42 PM Jeff Layton <jlayton@kernel.org> wrote:
> > > On Wed, 2020-04-29 at 11:46 +0200, Eduard Shishkin wrote:
> > > > On 4/28/20 2:23 PM, Jeff Layton wrote:
> > > > > On Mon, 2020-04-27 at 23:46 +0200, edward6@linux.ibm.com wrote:
> > > > > > From: Eduard Shishkin <edward6@linux.ibm.com>
> > > > > >
> > > > > > In the function handle_session() variable @features always
> > > > > > contains little endian order of bytes. Just because the feature
> > > > > > bits are packed bytewise from left to right in
> > > > > > encode_supported_features().
> > > > > >
> > > > > > However, test_bit(), called to check features availability, assumes
> > > > > > the host order of bytes in that variable. This leads to problems on
> > > > > > big endian architectures. Specifically it is impossible to mount
> > > > > > ceph volume on s390.
> > > > > >
> > > > > > This patch adds conversion from little endian to the host order
> > > > > > of bytes, thus fixing the problem.
> > > > > >
> > > > > > Signed-off-by: Eduard Shishkin <edward6@linux.ibm.com>
> > > > > > ---
> > > > > >   fs/ceph/mds_client.c | 4 ++--
> > > > > >   1 file changed, 2 insertions(+), 2 deletions(-)
> > > > > >
> > > > > > diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
> > > > > > index 486f91f..190598d 100644
> > > > > > --- a/fs/ceph/mds_client.c
> > > > > > +++ b/fs/ceph/mds_client.c
> > > > > > @@ -3252,7 +3252,7 @@ static void handle_session(struct ceph_mds_session *session,
> > > > > >           struct ceph_mds_session_head *h;
> > > > > >           u32 op;
> > > > > >           u64 seq;
> > > > > > - unsigned long features = 0;
> > > > > > + __le64 features = 0;
> > > > > >           int wake = 0;
> > > > > >           bool blacklisted = false;
> > > > > >
> > > > > > @@ -3301,7 +3301,7 @@ static void handle_session(struct ceph_mds_session *session,
> > > > > >                   if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
> > > > > >                           pr_info("mds%d reconnect success\n", session->s_mds);
> > > > > >                   session->s_state = CEPH_MDS_SESSION_OPEN;
> > > > > > -         session->s_features = features;
> > > > > > +         session->s_features = le64_to_cpu(features);
> > > > > >                   renewed_caps(mdsc, session, 0);
> > > > > >                   wake = 1;
> > > > > >                   if (mdsc->stopping)
> > > > >
> > > > > (cc'ing Zheng since he did the original patches here)
> > > > >
> > > > > Thanks Eduard. The problem is real, but I think we can just do the
> > > > > conversion during the decode.
> > > > >
> > > > > The feature mask words sent by the MDS are 64 bits, so if it's smaller
> > > > > we can assume that it's malformed. So, I don't think we need to handle
> > > > > the case where it's smaller than 8 bytes.
> > > > >
> > > > > How about this patch instead?
> > > >
> > > > Hi Jeff,
> > > >
> > > > This also works. Please, apply.
> > > >
> > > > Thanks,
> > > > Eduard.
> > > >
> > >
> > > Thanks. Merged into ceph-client/testing branch, and should make v5.8.
> >
> > I think this is stable material.  I'll tag it and get it queued up for 5.7-rc.
> >
> > Thanks,
> >
>
> Yeah, that sounds reasonable.
>
> If you're going to send up another PR, then we might want to add these
> bugfixes currently in the testing branch to it as well:
>
> 445645c8be5f fs/ceph:fix special error code in ceph_try_get_caps()
> 591681748b56 fs/ceph:fix double unlock in handle_cap_export()
> 0e84a1ebe161 ceph: ceph_kick_flushing_caps needs the s_mutex
>
> I'm not sure that any of them need to go to stable though. We might also
> want this one though:
>
> 7b3facb61440 ceph: reset i_requested_max_size if file write is not wanted
>
> ...but it'll probably need to be reworked due to merge conflicts if we
> move it ahead of some of the cap handling cleanup patches (or we could
> just pull those in too).
>
> Zheng, do you have an opinion here? Should 7b3facb61440 go to stable?

(dropping IBM folks)

"ceph: ceph_kick_flushing_caps needs the s_mutex" doesn't apply
either because cap_dirty list was moved.  I don't see the urgency on
these TBH.

The endianness fix and two error handling fixups from Wu are now
queued up.

Thanks,

                Ilya

Patch
diff mbox series

diff --git a/fs/ceph/mds_client.c b/fs/ceph/mds_client.c
index 486f91f..190598d 100644
--- a/fs/ceph/mds_client.c
+++ b/fs/ceph/mds_client.c
@@ -3252,7 +3252,7 @@  static void handle_session(struct ceph_mds_session *session,
 	struct ceph_mds_session_head *h;
 	u32 op;
 	u64 seq;
-	unsigned long features = 0;
+	__le64 features = 0;
 	int wake = 0;
 	bool blacklisted = false;
 
@@ -3301,7 +3301,7 @@  static void handle_session(struct ceph_mds_session *session,
 		if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
 			pr_info("mds%d reconnect success\n", session->s_mds);
 		session->s_state = CEPH_MDS_SESSION_OPEN;
-		session->s_features = features;
+		session->s_features = le64_to_cpu(features);
 		renewed_caps(mdsc, session, 0);
 		wake = 1;
 		if (mdsc->stopping)