Patch for reset in ini9100u [Initio 9100U(W)]
Hi.
I have an IWill 2935UW SCSI controller which uses the ini9100u driver. This has been working fine under 2.4 but I've recently built up a box of spare bits including the controller and installed 2.6 on it. The driver is marked broken in 2.6, apparently because of a lack of reset/abort functionality as it compiles and runs ok. So I've taken a stab at getting reset support back. Patch is below and it's received minimal testing - it boots, removes the callback trace and error message and doesn't seem to cause problems (the only disk in the machine is on this card).
-----------
drivers/scsi/ini9100u.c +13 -1
@@ -106,6 +106,8 @@* - Changed the assumption that HZ = 100 * 10/17/03 mc - v1.04 * - added new DMA API support +* 06/01/04 jmd - v1.04a +* - Re-add reset_bus support **************************************************************************/ #define CVT_LINUX_VERSION (V ,P ,S ) (V * 65536 + P * 256 + S)
@@ -149,6 +151,7 @@ .queuecommand = i91u_queue,// .abort = i91u_abort, // .reset = i91u_reset, + .eh_bus_reset_handler = i91u_bus_reset, .bios_param = i91u_biosparam, .can_queue = 1, .this_id = 1,
@@ -161,7 +164,7 @@char *i91uCopyright ="Copyright (C) 1996-98" ;char *i91uInitioName ="by Initio Corporation" ;char *i91uProductName ="INI-9X00U/UW" ; -char *i91uVersion ="v1.04" ; +char *i91uVersion ="v1.04a" ;#define TULSZ (sz) (sizeof(sz) / sizeof(sz[0]))#define TUL_RDWORD (x,y) (short)(inl((int)((ULONG)((ULONG)x+(UCHAR)y)) ))
@@ -587,6 +590,15 @@return tul_device_reset (pHCB, (ULONG ) SCpnt, SCpnt->device->id, reset_flags); } +int i91u_bus_reset (Scsi_Cmnd * SCpnt) +{ +HCS *pHCB; + + pHCB = (HCS *) SCpnt->device->host->base; +tul_reset_scsi_bus (pHCB); +return SUCCESS ; +} +/* * Return the "logical geometry" */
drivers/scsi/ini9100u.h +2 -1
@@ -82,10 +82,11 @@extern int i91u_queue (Scsi_Cmnd *,void (*done) (Scsi_Cmnd *));extern int i91u_abort (Scsi_Cmnd *);extern int i91u_reset (Scsi_Cmnd *,unsigned int ); +extern int i91u_bus_reset (Scsi_Cmnd *);extern int i91u_biosparam (struct scsi_device *,struct block_device *,sector_t ,int *); -#define i91u_REVID "Initio INI-9X00U/UW SCSI device driver; Revision: 1.03g" +#define i91u_REVID "Initio INI-9X00U/UW SCSI device driver; Revision: 1.04a"#define VIRT_TO_BUS (i) (unsigned int) virt_to_bus((void *)(i))#define ULONG unsigned long
-----------
J.
Signature · 4 lines
] http://www.earth.li/~noodles/ [] "0 tends to simplify things a bit [ ] PGP/GPG Key @ the.earth.li [] when you multiply by it..." -- [ ] via keyserver, web or email. [] Bill McColl [ ] RSA: 4DC4E7FD / DSA: 5B430367 [] [
Quoted 9 lines · Jonathan McDowell
I have an IWill 2935UW SCSI controller which uses the ini9100u driver. This has been working fine under 2.4 but I've recently built up a box of spare bits including the controller and installed 2.6 on it. The driver is marked broken in 2.6, apparently because of a lack of reset/abort functionality as it compiles and runs ok. So I've taken a stab at getting reset support back. Patch is below and it's received minimal testing - it boots, removes the callback trace and error message and doesn't seem to cause problems (the only disk in the machine is on this card).
I think it's a good beginning. However, there are some things that could be done to improve it.
drivers/scsi/ini9100u.c line 598
Quoted 6 lines
drivers/scsi/ini9100u.c +6
-0+int i91u_bus_reset (Scsi_Cmnd * SCpnt) +{ +HCS *pHCB; + + pHCB = (HCS *) SCpnt->device->host->base; +tul_reset_scsi_bus (pHCB);
This won't quite do beacuse tul_reset_scsi_bus() has some really nasty properties
Under the old error handler, the reset routine was responsible for resetting the bus, waiting the timeout (which tul_reset_scsi_bus() does with a busy wait) and flushing the queue.
In the new scheme, the eh thread takes care of all of this (including a nice thread based wait).
I think this may all work correctly if you change this call to:
tul_reset_scsi(pHCB, 0);
instead. That should simply reset the bus and not busy wait at all, which is really what the error handler is expecting.
James
Quoted 30 lines · James Bottomley
Quoted 9 lines · Jonathan McDowell
I have an IWill 2935UW SCSI controller which uses the ini9100u driver. This has been working fine under 2.4 but I've recently built up a box of spare bits including the controller and installed 2.6 on it. The driver is marked broken in 2.6, apparently because of a lack of reset/abort functionality as it compiles and runs ok. So I've taken a stab at getting reset support back. Patch is below and it's received minimal testing - it boots, removes the callback trace and error message and doesn't seem to cause problems (the only disk in the machine is on this card).
I think it's a good beginning. However, there are some things that could be done to improve it.
drivers/scsi/ini9100u.c line 598
Quoted 6 lines
drivers/scsi/ini9100u.c +6
-0+int i91u_bus_reset (Scsi_Cmnd * SCpnt) +{ +HCS *pHCB; + + pHCB = (HCS *) SCpnt->device->host->base; +tul_reset_scsi_bus (pHCB);This won't quite do beacuse
tul_reset_scsi_bus()has some really nasty propertiesUnder the old error handler, the reset routine was responsible for resetting the bus, waiting the timeout (which
tul_reset_scsi_bus()does with a busy wait) and flushing the queue.In the new scheme, the eh thread takes care of all of this (including a nice thread based wait).
Quoted 6 lines
I think this may all work correctly if you change this call to:
tul_reset_scsi(pHCB, 0);instead. That should simply reset the bus and not busy wait at all, which is really what the error handler is expecting.
Ahhh, cunning. I've changed this over and it seems to work ok. I'm not sure what sort of testing needs done though. As before it compiles and boots without problems and I haven't seen any issues in basic use.
J.
Signature · 4 lines
OK, if we can't have a tour, can | .''`. Debian GNU/Linux Developer
we at least have a look around? | : :' : Happy to accept PGP signed
| `. `' or encrypted mail - RSA +
| `- DSA keys on the keyservers.