LSI Logic 40919o fibre channel: scsi works ip not

LSI Logic 40919o fibre channel: scsi works ip not

[Mario Giammarco]

Hello,

I have two lsi logic 40919o 2gbit connected to a 2gbit switch.

They see hard disks but when I try to use them as ip card I obtain a partial 
failure: packets sometimes arrives sometimes no and on dmesg I see:

mptlan: ioc0/fc0: WARNING - IOC out of buckets! (priv->buckets_out =
126)
mptlan Mismatch between driver's buckets_out count and fw's
BucketsRemaining count has crossed the threshold, issuing a LanReset
to clear the fw's hashtable. You may want to check your
/var/log/messages for "CRC error" event notifications.
mptlan: ioc0/fc0: WARNING - IOC out of buckets! (priv->buckets_out =
125)
mptlan Mismatch between driver's buckets_out count and fw's
BucketsRemaining count has crossed the threshold, issuing a LanReset
to clear the fw's hashtable. You may want to check your
/var/log/messages for "CRC error" event notifications.
mptlan: ioc0/fc0: IOC says 1 buckets left. (priv->buckets_out = 126)
mptlan Mismatch between driver's buckets_out count and fw's
BucketsRemaining count has crossed the threshold, issuing a LanReset
to clear the fw's hashtable. You may want to check your
/var/log/messages for "CRC error" event notifications.
mptlan: ioc0/fc0: WARNING - IOC out of buckets! (priv->buckets_out =
126)


Boards are upgraded to latest bios I have tried several kernels.

Please tell me if you need other details.

Thanks in advance for any help.

Re: LSI Logic 40919o fibre channel: scsi works ip not

[Dan Aloni]

Mario Giammarco wrote:
> Hello,
>
> I have two lsi logic 40919o 2gbit connected to a 2gbit switch.
>
> They see hard disks but when I try to use them as ip card I obtain a partial 
> failure: packets sometimes arrives sometimes no and on dmesg I see:
>
> mptlan: ioc0/fc0: WARNING - IOC out of buckets! (priv->buckets_out =
> 126)
> mptlan Mismatch between driver's buckets_out count and fw's
> BucketsRemaining count has crossed the threshold, issuing a LanReset
> to clear the fw's hashtable. You may want to check your
> /var/log/messages for "CRC error" event notifications.
> mptlan: ioc0/fc0: WARNING - IOC out of buckets! (priv->buckets_out =
>
>   
AFAIK these messages occur as a result of bad frame tx/rx, and it
doesn't get handled by the hardware/firmware very well, and I'm
quite sure it never did.

Now regarding the whole thing surrounding mptlan, I don't think
that LSI officially supports that feature any more or willing to fix
any bugs for it in their firmware or driver. Is that right?

If so, we might as well remove that driver from the kernel.

-- 
Dan Aloni

Re: LSI Logic 40919o fibre channel: scsi works ip not

[Mario Giammarco]

Shirron, Stephen ha scritto:
> Hi Mario,
> 
> Can you tell me what firmware version you have on your 40919
> card(s)?  The easiest way is to "cat /proc/mpt/summary".

This is summary:
ioc0: LSIFC919, FwRev=02000f00h, Ports=1, 
MaxQ=1023,LanAddr=00:06:2B:07:FA:10, IRQ=209


> 
> How often do the errors occur?  dmesg provides no timestamps,
> while /var/log/messages does.
> 
Sorry but in messagese there is not this error. I can say you that the 
frequency of the error is proportional to data transferred. If I start a 
n ftp dmesg output goes crazy.

I have seen this in /var/log/messages (sorry no timestamps again):


Fusion MPT LAN driver 3.04.01
mptlan: ioc0: PortNum=0, ProtocolFlags=0Fh (ITLB)
mptlan: ioc0: Fusion MPT LAN device registered as 'fc0'
mptlan: ioc0/fc0: LanAddr = 00:06:2B:07:FA:10
mptlan: ioc0/fc0: interface up & active
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003

RE: LSI Logic 40919o fibre channel: scsi works ip not

[Shirron, Stephen]

Mario,

There appears to be a bug in the driver.

Go to mpt_lan_post_receive_buckets(), in mptlan.c.  Find:

		mf = mpt_get_msg_frame(LanCtx, mpt_dev);

Later find:

		pRecvReq = (LANReceivePostRequest_t *) mf;

Below that, should be:

		i = le16_to_cpu(mf->u.frame.hwhdr.msgctxu.fld.req_idx);
		mpt_dev->RequestNB[i] = 0;

The above two lines of code are present in some versions of mptctl.c
(the 2.06.xx versions and 3.02.xx versions) but are missing from some
others (3.03.xx, 3.04.xx, and 4.00.xx).

If you can add those two lines of code, rebuild mptctl.ko, and retest,
I believe that should take care of your reported problems.

Let me know, please.

stephen

-----Original Message-----
From: Mario Giammarco [mailto:mgiammarco-ml@4warding.com] 
Sent: Sunday, February 18, 2007 6:16 AM
To: Shirron, Stephen; linux-scsi@vger.kernel.org
Cc: Hickerson, Roger
Subject: Re: LSI Logic 40919o fibre channel: scsi works ip not

Shirron, Stephen ha scritto:
> Hi Mario,
> 
> Can you tell me what firmware version you have on your 40919
> card(s)?  The easiest way is to "cat /proc/mpt/summary".

This is summary:
ioc0: LSIFC919, FwRev=02000f00h, Ports=1, 
MaxQ=1023,LanAddr=00:06:2B:07:FA:10, IRQ=209


> 
> How often do the errors occur?  dmesg provides no timestamps,
> while /var/log/messages does.
> 
Sorry but in messagese there is not this error. I can say you that the 
frequency of the error is proportional to data transferred. If I start a

n ftp dmesg output goes crazy.

I have seen this in /var/log/messages (sorry no timestamps again):


Fusion MPT LAN driver 3.04.01
mptlan: ioc0: PortNum=0, ProtocolFlags=0Fh (ITLB)
mptlan: ioc0: Fusion MPT LAN device registered as 'fc0'
mptlan: ioc0/fc0: LanAddr = 00:06:2B:07:FA:10
mptlan: ioc0/fc0: interface up & active
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003
mptbase: ioc0: LogInfo(0x22000002): SubCl={LAN}
mptlan: ioc0/fc0: ERROR - Got a non-TURBO ReceivePostReply w/ 
PacketLength zero!
mptlan: MsgFlags = 80, IOCStatus = 8003

Re: LSI Logic 40919o fibre channel: scsi works ip not

[Mario Giammarco]

Il Monday 05 March 2007 20:46:10 Shirron, Stephen ha scritto:
> The code should look like this, in mpt_lan_post_receive_buckets():
>
> 			goto out;
> 		}
> 		pRecvReq = (LANReceivePostRequest_t *) mf;
>
> 		i = le16_to_cpu(mf->u.frame.hwhdr.msgctxu.fld.req_idx);
> 		mpt_dev->RequestNB[i] = 0;
>
> 		count = buckets;
> 		if (count > max)
> 			count = max;
>
> If you send me mptlan.c, I will edit it for you.  Then there cannot
> be a mistake.
>
> stephen
>
> -----Original Message-----
> From: Mario Giammarco [mailto:mgiammarco@gmail.com]
> Sent: Monday, March 05, 2007 2:39 PM
> To: Shirron, Stephen
> Cc: Mario Giammarco; Hickerson, Roger
> Subject: Re: LSI Logic 40919o fibre channel: scsi works ip not
>
> Sorry for delay replying. I have seen kernel 2.6.18 and 2.6.20 source
> and these lines are missinig:
>
>    i = le16_to_cpu(mf->u.frame.hwhdr.msgctxu.fld.req_idx);
>                 mpt_dev->RequestNB[i] = 0;
>
> I do not understand where to put them. Perhaps inside the for loop (i
> is the loop variable...) but I do not understand if I can choose a
> random point or I have to put it in a specific place.
>
> Thanks again for interest!
>
>
> 2007/2/16, Shirron, Stephen <Stephen.Shirron@lsi.com>:
> Mario,
>
> There appears to be a bug in the driver.
>
> Go to mpt_lan_post_receive_buckets(), in mptlan.c.  Find:
>
>                 mf = mpt_get_msg_frame(LanCtx, mpt_dev);
>
> Later find:
>
>                 pRecvReq = (LANReceivePostRequest_t *) mf;
>
> Below that, should be:
>
>                 i = le16_to_cpu(mf->u.frame.hwhdr.msgctxu.fld.req_idx);
>                 mpt_dev->RequestNB[i] = 0;
>
> The above two lines of code are present in some versions of mptctl.c
> (the 2.06.xx versions and 3.02.xx versions) but are missing from some
> others (3.03.xx, 3.04.xx, and 4.00.xx).
>
> If you can add those two lines of code, rebuild mptctl.ko, and retest,
> I believe that should take care of your reported problems.
>
> Let me know, please.
>
> stephen


Hello,
month ago I replied to you saying that this patch solved my bug.

Today I am trying kernel 2.6.22 and I see that it does not contain yet the 
patch.

Can you please put the patch in official driver?

Thank you very much!

Mario Giammarco

RE: LSI Logic 40919o fibre channel: scsi works ip not

[Shirron, Stephen]

Hi Mario,

Our internal driver kits have been fixed, but there has been a
delay in pushing this fix (along with other required changes)
out to kernel.org.  We are anticipating that the fix for this
issue will show up in kernel 2.6.24 (2.6.23 is essentially
closed).

stephen 

-----Original Message-----
From: Mario Giammarco [mailto:mgiammarco-ml@4warding.com] 
Sent: Thursday, September 20, 2007 11:34 AM
To: Shirron, Stephen
Cc: Mario Giammarco; Hickerson, Roger; linux-scsi@vger.kernel.org
Subject: Re: LSI Logic 40919o fibre channel: scsi works ip not

Il Monday 05 March 2007 20:46:10 Shirron, Stephen ha scritto:
> The code should look like this, in mpt_lan_post_receive_buckets():
>
> 			goto out;
> 		}
> 		pRecvReq = (LANReceivePostRequest_t *) mf;
>
> 		i = le16_to_cpu(mf->u.frame.hwhdr.msgctxu.fld.req_idx);
> 		mpt_dev->RequestNB[i] = 0;
>
> 		count = buckets;
> 		if (count > max)
> 			count = max;
>
> If you send me mptlan.c, I will edit it for you.  Then there cannot
> be a mistake.
>
> stephen
>
> -----Original Message-----
> From: Mario Giammarco [mailto:mgiammarco@gmail.com]
> Sent: Monday, March 05, 2007 2:39 PM
> To: Shirron, Stephen
> Cc: Mario Giammarco; Hickerson, Roger
> Subject: Re: LSI Logic 40919o fibre channel: scsi works ip not
>
> Sorry for delay replying. I have seen kernel 2.6.18 and 2.6.20 source
> and these lines are missinig:
>
>    i = le16_to_cpu(mf->u.frame.hwhdr.msgctxu.fld.req_idx);
>                 mpt_dev->RequestNB[i] = 0;
>
> I do not understand where to put them. Perhaps inside the for loop (i
> is the loop variable...) but I do not understand if I can choose a
> random point or I have to put it in a specific place.
>
> Thanks again for interest!
>
>
> 2007/2/16, Shirron, Stephen <Stephen.Shirron@lsi.com>:
> Mario,
>
> There appears to be a bug in the driver.
>
> Go to mpt_lan_post_receive_buckets(), in mptlan.c.  Find:
>
>                 mf = mpt_get_msg_frame(LanCtx, mpt_dev);
>
> Later find:
>
>                 pRecvReq = (LANReceivePostRequest_t *) mf;
>
> Below that, should be:
>
>                 i =
le16_to_cpu(mf->u.frame.hwhdr.msgctxu.fld.req_idx);
>                 mpt_dev->RequestNB[i] = 0;
>
> The above two lines of code are present in some versions of mptctl.c
> (the 2.06.xx versions and 3.02.xx versions) but are missing from some
> others (3.03.xx, 3.04.xx, and 4.00.xx).
>
> If you can add those two lines of code, rebuild mptctl.ko, and retest,
> I believe that should take care of your reported problems.
>
> Let me know, please.
>
> stephen


Hello,
month ago I replied to you saying that this patch solved my bug.

Today I am trying kernel 2.6.22 and I see that it does not contain yet
the 
patch.

Can you please put the patch in official driver?

Thank you very much!

Mario Giammarco

RE: LSI Logic 40919o fibre channel: scsi works ip not

[Moore, Eric]

On Saturday, February 17, 2007 9:04 AM,  Mario Giammarco wrote: 

> Now regarding the whole thing surrounding mptlan, I don't think
> that LSI officially supports that feature any more or willing to fix
> any bugs for it in their firmware or driver. Is that right?
> 
> If so, we might as well remove that driver from the kernel.
> 

No, don't be thinking about removing that driver. LSI Logic still
supports this driver.  I had forwarded Mario's original email to Stephen
Shirron last week. He should of been contacted by now, if not, please
let me know.


Eric Moore
LSI Logic

RE: LSI Logic 40919o fibre channel: scsi works ip not

[Moore, Eric]

On Saturday, February 17, 2007 9:04 AM,  Mario Giammarco wrote: 

> Now regarding the whole thing surrounding mptlan, I don't think
> that LSI officially supports that feature any more or willing to fix
> any bugs for it in their firmware or driver. Is that right?
> 
> If so, we might as well remove that driver from the kernel.
> 

No, don't be thinking about removing that driver. LSI Logic still
supports this driver.  I had forwarded Mario's original email to Stephen
Shirron last week. He should of been contacted by now, if not, please
let me know.


Eric Moore
LSI Logic