Re: [PATCH 2/2 linux-next] cifs: Make big endian multiplex ID sequences monotonic on the wire
On 10/16/2013 09:25 AM, Shirish Pargaonkar wrote: > On Wed, Oct 16, 2013 at 10:09 AM, Tim Gardner wrote: >> The multiplex identifier (MID) in the SMB header is only >> ever used by the client, in conjunction with PID, to match responses >> from the server. As such, the endianess of the MID is not important. >> However, When tracing packet sequences on the wire, protocol analyzers >> such as wireshark display MID as little endian. It is much more informative >> for the on-the-wire MID sequences to match debug information emitted by the >> CIFS driver. Therefore, one should write and read MID in the SMB header >> assuming it is always little endian. >> >> Observed from wireshark during the protocol negotiation >> and session setup: >> >> Multiplex ID: 256 >> Multiplex ID: 256 >> Multiplex ID: 512 >> Multiplex ID: 512 >> Multiplex ID: 768 >> Multiplex ID: 768 >> >> After this patch on-the-wire MID values begin at 1 and increase >> monotonically. >> >> Introduce get_next_mid64() for the internal consumers that use the full 64 >> bit >> multiplex identifier. >> >> Introduce the helpers get_mid() and compare_mid() to make the endian >> translation clear. >> >> Cc: Steve French >> Signed-off-by: Tim Gardner >> --- >> >> I'm looking at some of this code in excrutiating detail because I'm having >> trouble >> with a backport of CIFS from 3.9.7 on an embedded 2.6.31 powerpc. Its >> failing the RawNTLMSSP >> authentication and is almost certainly an endian issue. x86 on the same code >> base works >> quite well. Am I making a rash assumption that CIFS in 3.9 stable worked on >> big endian ? > > Do you have this commit fdf96a907c1fbb93c633e2b7ede3b8df26d6a4c0 ? > This might help if you do not have it in your code base. > Yep - I found that one pretty early on. It solved a powerpc issue with session startup to an x86 Linux CIFS server. The current failing session startup is to an Apple 10.8.5 (Lion) server (serverNOS=@(#)PROGRAM:smbd PROJECT:smbx-136.20.1). rtg -- Tim Gardner t...@tpi.com www.tpi.com OR 503-601-0234 x102 MT 406-443-5357 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
Re: [PATCH 2/2 linux-next] cifs: Make big endian multiplex ID sequences monotonic on the wire
On Wed, Oct 16, 2013 at 10:09 AM, Tim Gardner wrote:
> The multiplex identifier (MID) in the SMB header is only
> ever used by the client, in conjunction with PID, to match responses
> from the server. As such, the endianess of the MID is not important.
> However, When tracing packet sequences on the wire, protocol analyzers
> such as wireshark display MID as little endian. It is much more informative
> for the on-the-wire MID sequences to match debug information emitted by the
> CIFS driver. Therefore, one should write and read MID in the SMB header
> assuming it is always little endian.
>
> Observed from wireshark during the protocol negotiation
> and session setup:
>
> Multiplex ID: 256
> Multiplex ID: 256
> Multiplex ID: 512
> Multiplex ID: 512
> Multiplex ID: 768
> Multiplex ID: 768
>
> After this patch on-the-wire MID values begin at 1 and increase monotonically.
>
> Introduce get_next_mid64() for the internal consumers that use the full 64 bit
> multiplex identifier.
>
> Introduce the helpers get_mid() and compare_mid() to make the endian
> translation clear.
>
> Cc: Steve French
> Signed-off-by: Tim Gardner
> ---
>
> I'm looking at some of this code in excrutiating detail because I'm having
> trouble
> with a backport of CIFS from 3.9.7 on an embedded 2.6.31 powerpc. Its failing
> the RawNTLMSSP
> authentication and is almost certainly an endian issue. x86 on the same code
> base works
> quite well. Am I making a rash assumption that CIFS in 3.9 stable worked on
> big endian ?
Do you have this commit fdf96a907c1fbb93c633e2b7ede3b8df26d6a4c0 ?
This might help if you do not have it in your code base.
>
> fs/cifs/cifsglob.h | 25 -
> fs/cifs/misc.c |9 +
> fs/cifs/smb1ops.c |4 ++--
> fs/cifs/smb2transport.c |2 +-
> fs/cifs/transport.c |2 +-
> 5 files changed, 33 insertions(+), 9 deletions(-)
>
> diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
> index 52b6f6c..535e324 100644
> --- a/fs/cifs/cifsglob.h
> +++ b/fs/cifs/cifsglob.h
> @@ -620,11 +620,34 @@ set_credits(struct TCP_Server_Info *server, const int
> val)
> }
>
> static inline __u64
> -get_next_mid(struct TCP_Server_Info *server)
> +get_next_mid64(struct TCP_Server_Info *server)
> {
> return server->ops->get_next_mid(server);
> }
>
> +static inline __u16
> +get_next_mid(struct TCP_Server_Info *server)
> +{
> + __u16 mid = get_next_mid64(server);
> + /*
> +* The value in the SMB header should be little endian for easy
> +* on-the-wire decoding.
> +*/
> + return cpu_to_le16(mid);
> +}
> +
> +static inline __u16
> +get_mid(const struct smb_hdr *smb)
> +{
> + return le16_to_cpu(smb->Mid);
> +}
> +
> +static inline bool
> +compare_mid(__u16 mid, const struct smb_hdr *smb)
> +{
> + return mid == le16_to_cpu(smb->Mid);
> +}
> +
> /*
> * When the server supports very large reads and writes via POSIX extensions,
> * we can allow up to 2^24-1, minus the size of a READ/WRITE_AND_X header,
> not
> diff --git a/fs/cifs/misc.c b/fs/cifs/misc.c
> index 298e31e..c851d41 100644
> --- a/fs/cifs/misc.c
> +++ b/fs/cifs/misc.c
> @@ -295,7 +295,8 @@ check_smb_hdr(struct smb_hdr *smb)
> if (smb->Command == SMB_COM_LOCKING_ANDX)
> return 0;
>
> - cifs_dbg(VFS, "Server sent request, not response. mid=%u\n",
> smb->Mid);
> + cifs_dbg(VFS, "Server sent request, not response. mid=%u\n",
> +le16_to_cpu(smb->Mid));
> return 1;
> }
>
> @@ -358,11 +359,11 @@ checkSMB(char *buf, unsigned int total_read)
> return 0; /* bcc wrapped */
> }
> cifs_dbg(FYI, "Calculated size %u vs length %u mismatch for
> mid=%u\n",
> -clc_len, 4 + rfclen, smb->Mid);
> +clc_len, 4 + rfclen, le16_to_cpu(smb->Mid));
>
> if (4 + rfclen < clc_len) {
> cifs_dbg(VFS, "RFC1001 size %u smaller than SMB for
> mid=%u\n",
> -rfclen, smb->Mid);
> +rfclen, le16_to_cpu(smb->Mid));
> return -EIO;
> } else if (rfclen > clc_len + 512) {
> /*
> @@ -375,7 +376,7 @@ checkSMB(char *buf, unsigned int total_read)
> * data to 512 bytes.
> */
> cifs_dbg(VFS, "RFC1001 size %u more than 512 bytes
> larger than SMB for mid=%u\n",
> -rfclen, smb->Mid);
> +rfclen, le16_to_cpu(smb->Mid));
> return -EIO;
> }
> }
> diff --git a/fs/cifs/smb1ops.c b/fs/cifs/smb1ops.c
> index 8233b17..5598af9 100644
> --- a/fs/cifs/smb1ops.c
> +++ b/fs/cifs/smb1ops.c
> @@ -67,7 +67,7 @@ send_nt_cancel(struct TCP_Server_In
Re: [PATCH 2/2 linux-next] cifs: Make big endian multiplex ID sequences monotonic on the wire
On 10/16/2013 08:40 AM, Jeff Layton wrote:
> On Wed, 16 Oct 2013 09:09:50 -0600
> Tim Gardner wrote:
>
>> The multiplex identifier (MID) in the SMB header is only
>> ever used by the client, in conjunction with PID, to match responses
>> from the server. As such, the endianess of the MID is not important.
>> However, When tracing packet sequences on the wire, protocol analyzers
>> such as wireshark display MID as little endian. It is much more informative
>> for the on-the-wire MID sequences to match debug information emitted by the
>> CIFS driver. Therefore, one should write and read MID in the SMB header
>> assuming it is always little endian.
>>
>> Observed from wireshark during the protocol negotiation
>> and session setup:
>>
>> Multiplex ID: 256
>> Multiplex ID: 256
>> Multiplex ID: 512
>> Multiplex ID: 512
>> Multiplex ID: 768
>> Multiplex ID: 768
>>
>> After this patch on-the-wire MID values begin at 1 and increase
>> monotonically.
>>
>> Introduce get_next_mid64() for the internal consumers that use the full 64
>> bit
>> multiplex identifier.
>>
>> Introduce the helpers get_mid() and compare_mid() to make the endian
>> translation clear.
>>
>> Cc: Steve French
>> Signed-off-by: Tim Gardner
>> ---
>>
>> I'm looking at some of this code in excrutiating detail because I'm having
>> trouble
>> with a backport of CIFS from 3.9.7 on an embedded 2.6.31 powerpc. Its
>> failing the RawNTLMSSP
>> authentication and is almost certainly an endian issue. x86 on the same code
>> base works
>> quite well. Am I making a rash assumption that CIFS in 3.9 stable worked on
>> big endian ?
>>
>> fs/cifs/cifsglob.h | 25 -
>> fs/cifs/misc.c |9 +
>> fs/cifs/smb1ops.c |4 ++--
>> fs/cifs/smb2transport.c |2 +-
>> fs/cifs/transport.c |2 +-
>> 5 files changed, 33 insertions(+), 9 deletions(-)
>>
>> diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
>> index 52b6f6c..535e324 100644
>> --- a/fs/cifs/cifsglob.h
>> +++ b/fs/cifs/cifsglob.h
>> @@ -620,11 +620,34 @@ set_credits(struct TCP_Server_Info *server, const int
>> val)
>> }
>>
>> static inline __u64
>> -get_next_mid(struct TCP_Server_Info *server)
>> +get_next_mid64(struct TCP_Server_Info *server)
>> {
>> return server->ops->get_next_mid(server);
>> }
>>
>> +static inline __u16
>> +get_next_mid(struct TCP_Server_Info *server)
>> +{
>> +__u16 mid = get_next_mid64(server);
>> +/*
>> + * The value in the SMB header should be little endian for easy
>> + * on-the-wire decoding.
>> + */
>> +return cpu_to_le16(mid);
>> +}
>> +
>> +static inline __u16
>> +get_mid(const struct smb_hdr *smb)
>> +{
>> +return le16_to_cpu(smb->Mid);
>> +}
>> +
>> +static inline bool
>> +compare_mid(__u16 mid, const struct smb_hdr *smb)
>> +{
>> +return mid == le16_to_cpu(smb->Mid);
>> +}
>> +
>> /*
>> * When the server supports very large reads and writes via POSIX
>> extensions,
>> * we can allow up to 2^24-1, minus the size of a READ/WRITE_AND_X header,
>> not
>> diff --git a/fs/cifs/misc.c b/fs/cifs/misc.c
>> index 298e31e..c851d41 100644
>> --- a/fs/cifs/misc.c
>> +++ b/fs/cifs/misc.c
>> @@ -295,7 +295,8 @@ check_smb_hdr(struct smb_hdr *smb)
>> if (smb->Command == SMB_COM_LOCKING_ANDX)
>> return 0;
>>
>> -cifs_dbg(VFS, "Server sent request, not response. mid=%u\n", smb->Mid);
>> +cifs_dbg(VFS, "Server sent request, not response. mid=%u\n",
>> + le16_to_cpu(smb->Mid));
>> return 1;
>> }
>>
>> @@ -358,11 +359,11 @@ checkSMB(char *buf, unsigned int total_read)
>> return 0; /* bcc wrapped */
>> }
>> cifs_dbg(FYI, "Calculated size %u vs length %u mismatch for
>> mid=%u\n",
>> - clc_len, 4 + rfclen, smb->Mid);
>> + clc_len, 4 + rfclen, le16_to_cpu(smb->Mid));
>>
>> if (4 + rfclen < clc_len) {
>> cifs_dbg(VFS, "RFC1001 size %u smaller than SMB for
>> mid=%u\n",
>> - rfclen, smb->Mid);
>> + rfclen, le16_to_cpu(smb->Mid));
>> return -EIO;
>> } else if (rfclen > clc_len + 512) {
>> /*
>> @@ -375,7 +376,7 @@ checkSMB(char *buf, unsigned int total_read)
>> * data to 512 bytes.
>> */
>> cifs_dbg(VFS, "RFC1001 size %u more than 512 bytes
>> larger than SMB for mid=%u\n",
>> - rfclen, smb->Mid);
>> + rfclen, le16_to_cpu(smb->Mid));
>> return -EIO;
>> }
>> }
>
> It would be more efficient to byte-swap smb->Mid only once and store
> that in a u16 here and then print that out instead of doing it multiple
> times.
>
OK, I'll send out V2.
--
Tim Gardner t...@tpi.com www.tpi.com
O
Re: [PATCH 2/2 linux-next] cifs: Make big endian multiplex ID sequences monotonic on the wire
On Wed, 16 Oct 2013 09:09:50 -0600
Tim Gardner wrote:
> The multiplex identifier (MID) in the SMB header is only
> ever used by the client, in conjunction with PID, to match responses
> from the server. As such, the endianess of the MID is not important.
> However, When tracing packet sequences on the wire, protocol analyzers
> such as wireshark display MID as little endian. It is much more informative
> for the on-the-wire MID sequences to match debug information emitted by the
> CIFS driver. Therefore, one should write and read MID in the SMB header
> assuming it is always little endian.
>
> Observed from wireshark during the protocol negotiation
> and session setup:
>
> Multiplex ID: 256
> Multiplex ID: 256
> Multiplex ID: 512
> Multiplex ID: 512
> Multiplex ID: 768
> Multiplex ID: 768
>
> After this patch on-the-wire MID values begin at 1 and increase monotonically.
>
> Introduce get_next_mid64() for the internal consumers that use the full 64 bit
> multiplex identifier.
>
> Introduce the helpers get_mid() and compare_mid() to make the endian
> translation clear.
>
> Cc: Steve French
> Signed-off-by: Tim Gardner
> ---
>
> I'm looking at some of this code in excrutiating detail because I'm having
> trouble
> with a backport of CIFS from 3.9.7 on an embedded 2.6.31 powerpc. Its failing
> the RawNTLMSSP
> authentication and is almost certainly an endian issue. x86 on the same code
> base works
> quite well. Am I making a rash assumption that CIFS in 3.9 stable worked on
> big endian ?
>
> fs/cifs/cifsglob.h | 25 -
> fs/cifs/misc.c |9 +
> fs/cifs/smb1ops.c |4 ++--
> fs/cifs/smb2transport.c |2 +-
> fs/cifs/transport.c |2 +-
> 5 files changed, 33 insertions(+), 9 deletions(-)
>
> diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
> index 52b6f6c..535e324 100644
> --- a/fs/cifs/cifsglob.h
> +++ b/fs/cifs/cifsglob.h
> @@ -620,11 +620,34 @@ set_credits(struct TCP_Server_Info *server, const int
> val)
> }
>
> static inline __u64
> -get_next_mid(struct TCP_Server_Info *server)
> +get_next_mid64(struct TCP_Server_Info *server)
> {
> return server->ops->get_next_mid(server);
> }
>
> +static inline __u16
> +get_next_mid(struct TCP_Server_Info *server)
> +{
> + __u16 mid = get_next_mid64(server);
> + /*
> + * The value in the SMB header should be little endian for easy
> + * on-the-wire decoding.
> + */
> + return cpu_to_le16(mid);
> +}
> +
> +static inline __u16
> +get_mid(const struct smb_hdr *smb)
> +{
> + return le16_to_cpu(smb->Mid);
> +}
> +
> +static inline bool
> +compare_mid(__u16 mid, const struct smb_hdr *smb)
> +{
> + return mid == le16_to_cpu(smb->Mid);
> +}
> +
> /*
> * When the server supports very large reads and writes via POSIX extensions,
> * we can allow up to 2^24-1, minus the size of a READ/WRITE_AND_X header,
> not
> diff --git a/fs/cifs/misc.c b/fs/cifs/misc.c
> index 298e31e..c851d41 100644
> --- a/fs/cifs/misc.c
> +++ b/fs/cifs/misc.c
> @@ -295,7 +295,8 @@ check_smb_hdr(struct smb_hdr *smb)
> if (smb->Command == SMB_COM_LOCKING_ANDX)
> return 0;
>
> - cifs_dbg(VFS, "Server sent request, not response. mid=%u\n", smb->Mid);
> + cifs_dbg(VFS, "Server sent request, not response. mid=%u\n",
> + le16_to_cpu(smb->Mid));
> return 1;
> }
>
> @@ -358,11 +359,11 @@ checkSMB(char *buf, unsigned int total_read)
> return 0; /* bcc wrapped */
> }
> cifs_dbg(FYI, "Calculated size %u vs length %u mismatch for
> mid=%u\n",
> - clc_len, 4 + rfclen, smb->Mid);
> + clc_len, 4 + rfclen, le16_to_cpu(smb->Mid));
>
> if (4 + rfclen < clc_len) {
> cifs_dbg(VFS, "RFC1001 size %u smaller than SMB for
> mid=%u\n",
> - rfclen, smb->Mid);
> + rfclen, le16_to_cpu(smb->Mid));
> return -EIO;
> } else if (rfclen > clc_len + 512) {
> /*
> @@ -375,7 +376,7 @@ checkSMB(char *buf, unsigned int total_read)
>* data to 512 bytes.
>*/
> cifs_dbg(VFS, "RFC1001 size %u more than 512 bytes
> larger than SMB for mid=%u\n",
> - rfclen, smb->Mid);
> + rfclen, le16_to_cpu(smb->Mid));
> return -EIO;
> }
> }
It would be more efficient to byte-swap smb->Mid only once and store
that in a u16 here and then print that out instead of doing it multiple
times.
> diff --git a/fs/cifs/smb1ops.c b/fs/cifs/smb1ops.c
> index 8233b17..5598af9 100644
> --- a/fs/cifs/smb1ops.c
> +++ b/fs/cifs/smb1ops.c
> @@ -67,7 +67,7 @@ send_nt_cancel(struct TCP_Server_Info *server, void *buf,
>
[PATCH 2/2 linux-next] cifs: Make big endian multiplex ID sequences monotonic on the wire
The multiplex identifier (MID) in the SMB header is only
ever used by the client, in conjunction with PID, to match responses
from the server. As such, the endianess of the MID is not important.
However, When tracing packet sequences on the wire, protocol analyzers
such as wireshark display MID as little endian. It is much more informative
for the on-the-wire MID sequences to match debug information emitted by the
CIFS driver. Therefore, one should write and read MID in the SMB header
assuming it is always little endian.
Observed from wireshark during the protocol negotiation
and session setup:
Multiplex ID: 256
Multiplex ID: 256
Multiplex ID: 512
Multiplex ID: 512
Multiplex ID: 768
Multiplex ID: 768
After this patch on-the-wire MID values begin at 1 and increase monotonically.
Introduce get_next_mid64() for the internal consumers that use the full 64 bit
multiplex identifier.
Introduce the helpers get_mid() and compare_mid() to make the endian
translation clear.
Cc: Steve French
Signed-off-by: Tim Gardner
---
I'm looking at some of this code in excrutiating detail because I'm having
trouble
with a backport of CIFS from 3.9.7 on an embedded 2.6.31 powerpc. Its failing
the RawNTLMSSP
authentication and is almost certainly an endian issue. x86 on the same code
base works
quite well. Am I making a rash assumption that CIFS in 3.9 stable worked on big
endian ?
fs/cifs/cifsglob.h | 25 -
fs/cifs/misc.c |9 +
fs/cifs/smb1ops.c |4 ++--
fs/cifs/smb2transport.c |2 +-
fs/cifs/transport.c |2 +-
5 files changed, 33 insertions(+), 9 deletions(-)
diff --git a/fs/cifs/cifsglob.h b/fs/cifs/cifsglob.h
index 52b6f6c..535e324 100644
--- a/fs/cifs/cifsglob.h
+++ b/fs/cifs/cifsglob.h
@@ -620,11 +620,34 @@ set_credits(struct TCP_Server_Info *server, const int val)
}
static inline __u64
-get_next_mid(struct TCP_Server_Info *server)
+get_next_mid64(struct TCP_Server_Info *server)
{
return server->ops->get_next_mid(server);
}
+static inline __u16
+get_next_mid(struct TCP_Server_Info *server)
+{
+ __u16 mid = get_next_mid64(server);
+ /*
+* The value in the SMB header should be little endian for easy
+* on-the-wire decoding.
+*/
+ return cpu_to_le16(mid);
+}
+
+static inline __u16
+get_mid(const struct smb_hdr *smb)
+{
+ return le16_to_cpu(smb->Mid);
+}
+
+static inline bool
+compare_mid(__u16 mid, const struct smb_hdr *smb)
+{
+ return mid == le16_to_cpu(smb->Mid);
+}
+
/*
* When the server supports very large reads and writes via POSIX extensions,
* we can allow up to 2^24-1, minus the size of a READ/WRITE_AND_X header, not
diff --git a/fs/cifs/misc.c b/fs/cifs/misc.c
index 298e31e..c851d41 100644
--- a/fs/cifs/misc.c
+++ b/fs/cifs/misc.c
@@ -295,7 +295,8 @@ check_smb_hdr(struct smb_hdr *smb)
if (smb->Command == SMB_COM_LOCKING_ANDX)
return 0;
- cifs_dbg(VFS, "Server sent request, not response. mid=%u\n", smb->Mid);
+ cifs_dbg(VFS, "Server sent request, not response. mid=%u\n",
+le16_to_cpu(smb->Mid));
return 1;
}
@@ -358,11 +359,11 @@ checkSMB(char *buf, unsigned int total_read)
return 0; /* bcc wrapped */
}
cifs_dbg(FYI, "Calculated size %u vs length %u mismatch for
mid=%u\n",
-clc_len, 4 + rfclen, smb->Mid);
+clc_len, 4 + rfclen, le16_to_cpu(smb->Mid));
if (4 + rfclen < clc_len) {
cifs_dbg(VFS, "RFC1001 size %u smaller than SMB for
mid=%u\n",
-rfclen, smb->Mid);
+rfclen, le16_to_cpu(smb->Mid));
return -EIO;
} else if (rfclen > clc_len + 512) {
/*
@@ -375,7 +376,7 @@ checkSMB(char *buf, unsigned int total_read)
* data to 512 bytes.
*/
cifs_dbg(VFS, "RFC1001 size %u more than 512 bytes
larger than SMB for mid=%u\n",
-rfclen, smb->Mid);
+rfclen, le16_to_cpu(smb->Mid));
return -EIO;
}
}
diff --git a/fs/cifs/smb1ops.c b/fs/cifs/smb1ops.c
index 8233b17..5598af9 100644
--- a/fs/cifs/smb1ops.c
+++ b/fs/cifs/smb1ops.c
@@ -67,7 +67,7 @@ send_nt_cancel(struct TCP_Server_Info *server, void *buf,
mutex_unlock(&server->srv_mutex);
cifs_dbg(FYI, "issued NT_CANCEL for mid %u, rc = %d\n",
-in_buf->Mid, rc);
+le16_to_cpu(in_buf->Mid), rc);
return rc;
}
@@ -101,7 +101,7 @@ cifs_find_mid(struct TCP_Server_Info *server, char *buffer)
spin_lock(&GlobalMid_Lock);
list_for_each_entry(mid, &server->pending_mid_q, qhead) {
-
