SCSI -rc fixes for 2.6.10-rc3

This one is another set of small driver fixes

It is available from:

bk://linux-scsi.bkbits.net/scsi-for-linus-2.6

The short changelog is:

Andreas Herrmann:
o zfcp: act enhancements corrections

Castor Fu:
o support for 3PAR sparse SCSI LUNs

Douglas Gilbert:
o extracting resid from struct scsi_cmnd

Eric Moore:
o mptfusion: kill fusion init called
o mptfusion: resid cleanup
o mptfusion: replace chip_type
o mptfusion: streamline queuecommand
o mptfusion: streamline slave_alloc
o mptfusion: delete MPTSCSIH_DBG_TIMEOUT

James Bottomley:
o aic7xxx: fix compiler warning from dma mask assignement
o mptfusion: command line parameters

Mark Haverkamp:
o 2.6.9 aacraid: Support ROMB RAID/SCSI mode - Resend -

Matthew Wilcox:
o Blacklist devices that falsely claim an echo buffer
o SCSI: spi_transport set_signalling is buggy

Mike Miller:
o cciss: cciss_ioctl return code fix

And the diffstat is:

 block/cciss.c                  |    2 
 message/fusion/mptbase.c       |   83 +----
 message/fusion/mptbase.h       |   36 --
 message/fusion/mptctl.c        |    8 
 message/fusion/mptscsih.c      |  633 ++++++++++-------------------------------
 message/fusion/mptscsih.h      |   11 
 s390/scsi/zfcp_def.h           |    4 
 s390/scsi/zfcp_erp.c           |   38 +-
 s390/scsi/zfcp_ext.h           |    4 
 s390/scsi/zfcp_fsf.c           |   53 ---
 s390/scsi/zfcp_scsi.c          |    3 
 scsi/Kconfig                   |    2 
 scsi/aacraid/aachba.c          |   46 ++
 scsi/aacraid/aacraid.h         |    3 
 scsi/aic7xxx/aic79xx_osm.h     |    2 
 scsi/aic7xxx/aic79xx_osm_pci.c |   14 
 scsi/aic7xxx/aic7xxx_osm.h     |    2 
 scsi/aic7xxx/aic7xxx_osm_pci.c |    7 
 scsi/scsi_devinfo.c            |    1 
 scsi/scsi_lib.c                |    2 
 scsi/scsi_transport_spi.c      |    9 
 21 files changed, 292 insertions(+), 671 deletions(-)

James

Here is one more. Currently imm uses page_address() which can crash on highmem. It's not directly doable to map the pages properly, at least not without changing some code. In lack of a ->bounce_highio member in the scsi host template, just set ->unchecked_isa_dma which will just bounce everything for us. imm isn't performance critical by any stretch of the imagination, so...

Usually I'd not encourage such a silly hack, but in lack of hardware for testing (who has it??), this should suffice as it is obviously correct.

===== drivers/scsi/imm.c 1.39 vs edited =====

@@ -1140,6 +1140,10 @@
 	.use_clustering		= ENABLE_CLUSTERING,
 	.can_queue		= 1,
 	.slave_alloc		= imm_adjust_queue,
+	.unchecked_isa_dma	= 1, /* imm cannot deal with highmem, so
+				      * this is an easy trick to ensure
+				      * all io pages for this host reside
+				      * in low memory */
 };
 
 /***************************************************************************

Jens Axboe

I have a drive, but I'm not near a highmem machine atm. I can test it next week if no one else does so.

~Randy

So would you be amenable to fixing it properly? It looks like what needs to happen is that it needs to accept its commands into the scsi_pointer with page and offset (scsi_pointer doesn't have these fields, but you could probably cast in ptr and dma_handle). Then do a kmap/kunmap around imm_in() and imm_out() when it's operating on SCp.ptr.

James

Yep, that would be the optimal. But still, given the hardware, I feel the hacky workaround would be acceptable. If Randy wants to fix it properly being able to test it, perfect.

Jens Axboe

Actually, I already put your patch in the scsi-rc-fixes-2.6 tree since it was obviously correct. I was just being hopeful ...

James