Skip to contentSkip to conversation

sunrpc: fix xs_read_xdr_buf for partial pages receive

PATCH Superseded Dan Aloni 2 participants 4 messages Dec 4, 2020 – Dec 5, 2020

Subsystems (3)
lore
Share & tools
lore

Dan Aloni

author
#0 lore raw

PATCH 1/1 sunrpc: fix xs_read_xdr_buf for partial pages receive

When receiving pages data, return value 'ret' when positive includes buf->page_base, so we should subtract it similarly to how it is done for offset.

This was discovered on the very rare cases where the server returned a chunk of bytes that when added to the already received amount of bytes for the pages happened to match the current recv.len, for example on this case:

     buf->page_base : 258356
     actually received from socket: 1740
     ret : 260096
     want : 260096

In this case neither of the two 'if ... goto out' trigger, and we continue to tail parsing.

Worth to mention that the ensuing EMSGSIZE from the continued execution of xs_read_xdr_buf may be observed by an application due to 4 superfluous bytes being added to the pages data.

Fixes 277e4ab7d53 ("SUNRPC: Simplify TCP receive code by switching to

using iterators")

---
 net/sunrpc/xprtsock.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/sunrpc/xprtsock.c +1 -1
index 7090bbee0ec5..42b680a60d38 100644
@@ -436,7 +436,7 @@ xs_read_xdr_buf(struct socket *sock, struct msghdr *msg, int flags,
 		offset += ret - buf->page_base;
 		if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
 			goto out;
-		if (ret != want)
+		if (ret - buf->page_base != want)
 			goto out;
 		seek = 0;
 	} else {
Signature · 1 line
2.26.2
#1 lore raw
Quoted 41 lines · Dan Aloni

When receiving pages data, return value 'ret' when positive includes buf->page_base, so we should subtract it similarly to how it is done
for offset.

This was discovered on the very rare cases where the server returned a
chunk of bytes that when added to the already received amount of bytes
for the pages happened to match the current recv.len, for example on this case:

     buf->page_base : 258356
     actually received from socket: 1740
     ret : 260096
     want : 260096

In this case neither of the two 'if ... goto out' trigger, and we continue to tail parsing.

Worth to mention that the ensuing EMSGSIZE from the continued execution of
xs_read_xdr_buf may be observed by an application due to 4 superfluous
bytes being added to the pages data.

Fixes 277e4ab7d53 ("SUNRPC: Simplify TCP receive code by switching

to
using iterators")

---
 net/sunrpc/xprtsock.c | 2 +-

 1 file changed, 1 insertion(+), 1 deletion(-)
net/sunrpc/xprtsock.c +0 -0
index 7090bbee0ec5..42b680a60d38 100644
@@ -436,7 +436,7 @@ xs_read_xdr_buf(struct socket *sock, struct
 msghdr *msg, int flags,
                 offset += ret - buf->page_base;
Quoted 8 lines

                if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
                        goto out;
-               if (ret != want)
+               if (ret - buf->page_base != want) goto out;
                seek = 0;
        } else {

Ouch... Well spotted!

Hmm... I think we want to just subtract out the buf->page_base from the value of 'ret' after we call xs_flush_bvec() and then adjust the calculation of 'offset' in the next line. That's more efficient.

Can you please resend with that change?

Signature · 3 lines
Trond Myklebust
Linux NFS client maintainer, Hammerspace
[email protected]
author
#2 lore raw

PATCH v2 1/1 sunrpc: fix xs_read_xdr_buf for partial pages receive

When receiving pages data, return value 'ret' when positive includes buf->page_base, so we should subtract that before it is used for changing offset and comparing against want.

This was discovered on the very rare cases where the server returned a chunk of bytes that when added to the already received amount of bytes for the pages happened to match the current recv.len, for example on this case:

     buf->page_base : 258356
     actually received from socket: 1740
     ret : 260096
     want : 260096

In this case neither of the two 'if ... goto out' trigger, and we continue to tail parsing.

Worth to mention that the ensuing EMSGSIZE from the continued execution of xs_read_xdr_buf may be observed by an application due to 4 superfluous bytes being added to the pages data.

Fixes 277e4ab7d53 ("SUNRPC: Simplify TCP receive code by switching to

using iterators")

---
 net/sunrpc/xprtsock.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
net/sunrpc/xprtsock.c +2 -1
index 7090bbee0ec5..9a4300f07e4c 100644
@@ -433,7 +433,8 @@ xs_read_xdr_buf(struct socket *sock, struct msghdr *msg, int flags,
 		if (ret <= 0)
 			goto sock_err;
 		xs_flush_bvec(buf->bvec, ret, seek + buf->page_base);
-		offset += ret - buf->page_base;
+		ret -= buf->page_base;
+		offset += ret;
 		if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
 			goto out;
 		if (ret != want)
Signature · 1 line
2.26.2
author
#3 lore raw
In reply to Trond Myklebust
Quoted 1 line · Trond Myklebust

On Fri, 2020-12-04 at 20:34 +0200, Dan Aloni wrote:

[..]

Quoted 16 lines
Quoted 5 lines
net/sunrpc/xprtsock.c +0 -0
@@ -436,7 +436,7 @@ xs_read_xdr_buf(struct socket *sock, struct
 msghdr *msg, int flags,
                 offset += ret - buf->page_base;
Quoted 8 lines

                if (offset == count || msg->msg_flags & (MSG_EOR|MSG_TRUNC))
                        goto out;
-               if (ret != want)
+               if (ret - buf->page_base != want) goto out;
                seek = 0;
        } else {

Ouch... Well spotted!

Quoted 3 lines

Hmm... I think we want to just subtract out the buf->page_base from the value of 'ret' after we call xs_flush_bvec() and then adjust the calculation of 'offset' in the next line. That's more efficient.

Yes, it works out, though after being aware that the positive value of ret when returned from xs_read_xdr_buf and propagated upward has no effect except on traces.

Signature · 1 line
Dan Aloni