URL
https://opencores.org/ocsvn/or1k/or1k/trunk
Subversion Repositories or1k
[/] [or1k/] [trunk/] [rc203soc/] [sw/] [uClinux/] [drivers/] [scsi/] [ChangeLog.ncr53c8xx] - Rev 1765
Compare with Previous | Blame | View Log
Sun Dec 13 18:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.1e
- Same work-around as for the 53c876 rev <= 0x15 for 53c896 rev 1:
Disable overlapped arbitration. This will not make difference
since the chip has on-chip RAM.
Thu Nov 26 22:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.1d
- The SISL RAID change requires now remap_pci_mem() stuff to be
compiled for __i386__ when normal IOs are used.
- Minor spelling fixes in doc files.
Sat Nov 21 18:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.1c
- Ignore chips that are driven by SISL RAID (DAC 960).
Change sent by Leonard Zubkoff and slightly reworked.
- Still a buglet in the tags initial settings that needed to be fixed.
It was not possible to disable TGQ at system startup for devices
that claim TGQ support. The driver used at least 2 for the queue
depth but did'nt keep track of user settings for tags depth lower
than 2.
Wed Nov 11 10:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.1b
- The driver was unhappy when configured with default_tags > MAX_TAGS
Hopefully doubly-fixed.
- Update the Configure.help driver section that speaks of TAGS.
Wed Oct 21 21:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.1a
- Changes from Eddie Dost for Sparc and Alpha:
ioremap/iounmap support for Sparc.
pcivtophys changed to bus_dvma_to_phys.
- Add the 53c876 description to the chip table. This is only usefull
for printing the right name of the controller.
- DEL-441 Item 2 work-around for the 53c876 rev <= 5 (0x15).
- Add additionnal checking of INQUIRY data:
Check INQUIRY data received length is at least 7. Byte 7 of
inquiry data contains device features bits and the driver might
be confused by garbage. Also check peripheral qualifier.
- Cleanup of the SCSI tasks management:
Remove the special case for 32 tags. Now the driver only uses the
scheme that allows up to 64 tags per LUN.
Merge some code from the 896 driver.
Use a 1,3,5,...MAXTAGS*2+1 tag numbering. Previous driver could
use any tag number from 1 to 253 and some non conformant devices
might have problems with large tag numbers.
- 'no_sync' changed to 'no_disc' in the README file. This is an old
and trivial mistake that seems to demonstrate the README file is
not often read. :)
Sun Oct 4 14:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.0i
- Cosmetic changes for sparc (but not for the driver) that needs
__irq_itoa() to be used for printed IRQ value to be understandable.
- Some problems with the driver that didn't occur using driver 2.5f
were due to a SCSI selection problem triggered by a clearly
documented feature that in fact seems not to work: (53C8XX chips
are claimed by the manuals to be able to execute SCSI scripts just
after abitration while the SCSI core is performing SCSI selection).
This optimization is broken and has been removed.
- Some broken scsi devices are confused when a negotiation is started
on a LUN that does not correspond to a real device. According to
SCSI specs, this is a device firmware bug. This has been worked
around by only starting negotiation if the LUN has previously be
used for at least 1 successful SCSI command.
- The 'last message sent' printed out on M_REJECT message reception
was read from the SFBR i/o register after the previous message had
been sent.
This was not correct and affects all previous driver versions and
the original FreeBSD one as well. The SCSI scripts has been fixed
so that it now provides the right information to the C code.
Sat Jul 18 13:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.0g
- Preliminary fixes for Big Endian (sent by Eddie C. Dost).
Big Endian architectures should work again with the driver.
Eddie's patch has been partially applied since current 2.1.109
does not have all the Sparc changes of the vger tree.
- Use of BITS_PER_LONG instead of (~0UL == 0xffffffffUL) has fixed
the problem observed when the driver was compiled using EGCS or
PGCC.
Mon Jul 13 20:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.0f
- Some spelling fixes.
- linux/config.h misplaced in ncr53c8xx.h
- MODULE_PARM stuff added for linux 2.1.
- check INQUIRY response data format is exactly 2.
- use BITS_PER_LONG if defined.
Sun Jun 28 12:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.0e
- Some cleanup, spelling fixes, version checks, documentations
changes, etc ...
Sat Jun 20 20:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.0c
- Add a boot setup option that allows to set up device queue depths
at boot-up. This option is very usefull since Linux does not
allow to change scsi device queue depth once the system has been
booted up.
Sun Jun 15 23:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.0a
- Support for up to 64 TAGS per LUN.
- Rewrite the TARGET vs LUN capabilities management.
CmdQueue is now handled as a LUN capability as it shall be.
This also fixes a bug triggered when disabling tagged command
queuing for a device that had this feature enabled.
- Remove the ncr_opennings() stuff that was useless under Linux
and hard to understand to me.
- Add "setverbose" procfs driver command. It allows to tune
verbose level after boot-up. Setting this level to zero, for
example avoid flooding the syslog file.
- Add KERN_XXX to some printk's.
Tue Jun 10 23:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 3.0
- Linux config changes for 2.0.34:
Remove NVRAM detection config option. This option is now enabled
by default but can be disabled by editing the driver header file.
Add a PROFILE config option.
- Update Configure.help
- Add calls to new function mdelay() for milli-seconds delay if
kernel version >= 2.1.105.
- Replace all printf(s) by printk(s). After all, the ncr53c8xx is
a driver for Linux.
- Perform auto-sense on COMMAND TERMINATED. Not sure it is usefull.
- Some other minor changes.
Tue Jun 4 23:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6n
- Code cleanup and simplification:
Remove kernel 1.2.X and 1.3.X support.
Remove the _old_ target capabilities table.
Remove the error recovery code that have'nt been really usefull.
Use a single alignment boundary (CACHE_LINE_SIZE) for data
structures.
- Several aggressive SCRIPTS optimizations and changes:
Reselect SCRIPTS code rewritten.
Support for selection/reselection without ATN.
And some others.
- Miscallaneous changes in the C code:
Count actual number of CCB queued to the controller (future use).
Lots of other minor changes.
Wed May 13 20:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6m
- Problem of missed SCSI bus reset with the 53C895 fixed by
Richard Waltham. The 53C895 needs about 650 us for the bus
mode to settle. Delays used while resetting the controller
and the bus have been adjusted. Thanks Richard!
- Some simplification for 64 bit arch done ccb address testing.
- Add a check of the MSG_OUT phase after Selection with ATN.
- The new tagged queue stuff seems ok, so some informationnal
message have been conditionned by verbose >= 3.
- Donnot reset if a SBMC interrupt reports the same bus mode.
- Print out the whole driver set-up. Some options were missing and
the print statement was misplaced for modules.
- Ignore a SCSI parity interrupt if the chip is not connected to
the SCSI bus.
Sat May 1 16:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6l
- Add CCB done queue support for Alpha and perhaps some other
architectures.
- Add some barriers to enforce memory ordering for x86 and
Alpha architectures.
- Fix something that looks like an old bug in the nego SIR
interrupt code in case of negotiation failure.
Sat Apr 25 21:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6k
- Remove all accesses to the on-chip RAM from the C code:
Use SCRIPTS to load the on-chip RAM.
Use SCRIPTS to repair the start queue on selection timeout.
Use the copy of script in main memory to calculate the chip
context on phase mismatch.
- The above allows now to use the on-chip RAM without requiring
to get access to the on-chip RAM from the C code. This makes
on-chip RAM useable for linux-1.2.13 and for Linux-Alpha for
instance.
- Some simplifications and cleanups in the SCRIPTS and C code.
- Buglet fixed in parity error recovery SCRIPTS (never tested).
- Minor updates in README.ncr53c8xx.
Wed Apr 15 21:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6j
- Incorporate changes from linux-2.1.95 ncr53c8xx driver version.
- Add SMP support for linux-2.1.95 and above.
- Fix a bug when QUEUE FULL is returned and no commands are
disconnected. This happens with Atlas I / L912 and may happen
with Atlas II / LXY4.
- Nail another one on CHECK condition when requeuing the command
for auto-sense.
- Call scsi_done() for all completed commands after interrupt
handling.
- Increase the done queue to 24 entries.
Sat Apr 4 20:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6i
- CTEST0 is used by the 53C885 for Power Management and
priority setting between the 2 functions.
Use SDID instead as actual target number. Just have had to
overwrite it with SSID on reselection.
- Split DATA_IN and DATA_OUT scripts into 2 sub-scripts.
64 segments are moved from on-chip RAM scripts.
If more segments, a script in main memory is used for the
additionnal segments.
- Since the SCRIPTS processor continues SCRIPTS execution after
having won arbitration, do some stuff prior to testing any SCSI
phase on reselection. This should have the vertue to process
scripts in parallel with the SCSI core performing selection.
- Increase the done queue to 12 entries.
Sun Mar 29 12:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6h
- Some fixes.
Tue Mar 26 23:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6g
- New done queue. 8 entries by default (6 always useable).
Can be increased if needed.
- Resources management using doubly linked queues.
- New auto-sense and QUEUE FULL handling that does not need to
stall the NCR queue any more.
- New CCB starvation avoiding algorithm.
- Prepare CCBs for SCSI commands that cannot be queued, instead of
inserting these commands into the waiting list. The waiting list
is now only used while resetting and when memory for CCBs is not
yet available?
Sun Feb 8 22:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6f
- Some fixes in order to really support the 53C895, at least with
FAST-20 devices.
- Heavy changes in the target/lun resources management to allow
the scripts to jump directly to the CCB on reselection instead
of walking on the lun CCBs list. Up to 32 tags per lun are now
supported without script processor and PCI traffic overhead.
Sun Jan 11 22:00 1998 Gerard Roudier (groudier@club-internet.fr)
* revision 2.6d
- new (different ?) implementation of the start queue:
Use a simple CALL to a launch script in the CCB.
- implement a minimal done queue (1 entry :-) ).
this avoid scanning all CCBs on INT FLY (Only scan all CCBs, on
overflow). Hit ratio is better than 99.9 % on my system, so no
need to have a larger done queue.
- generalization of the restart of CCB on special condition as
Abort, QUEUE FULL, CHECK CONDITION.
This has been called 'silly scheduler'.
- make all the profiling code conditionned by a config option.
This spare some PCI traffic and C code when this feature is not
needed.
- handle more cleanly the situation where direction is unknown.
The pointers patching is now performed by the SCRIPTS processor.
- remove some useless scripts instructions.
Ported from driver 2.5 series:
------------------------------
- Use FAST-5 instead of SLOW for slow scsi devices according to
new SPI-2 draft.
- Make some changes in order to accomodate with 875 rev <= 3
device errata listing 397. Minor consequences are:
. Leave use of PCI Write and Invalidate under user control.
Now, by default the driver does not enable PCI MWI and option
'specf:y' is required in order to enable this feature.
. Memory Read Line is not enabled for 875 and 875-like chips.
. Programmed burst length set to 64 DWORDS (instead of 128).
(Note: SYMBIOS uses 32 DWORDS for the SDMS BIOS)
- Add 'buschk' boot option.
This option enables checking of SCSI BUS data lines after SCSI
RESET (set by default). (Submitted by Richard Waltham).
- Update the README file.
- Dispatch CONDITION MET and RESERVATION CONFLICT scsi status
as OK driver status.
- Update the README file and the Symbios NVRAM format definition
with removable media flags values (available with SDMS 4.09).
- Several PCI configuration registers fix-ups for powerpc.
(Patch sent by Cort).