145 lines
5.9 KiB
Plaintext
145 lines
5.9 KiB
Plaintext
Sat Dec 30 21:30 2000 Gerard Roudier
|
|
* version sym-2.1.0-20001230
|
|
- Initial release of SYM-2.
|
|
|
|
Mon Jan 08 21:30 2001 Gerard Roudier
|
|
* version sym-2.1.1-20010108
|
|
- Change a couple of defines containing ncr or NCR by their
|
|
equivalent containing sym or SYM instead.
|
|
|
|
Sun Jan 14 22:30 2001 Gerard Roudier
|
|
* version sym-2.1.2-20010114
|
|
- Fix a couple of printfs:
|
|
* Add the target number to the display of transfer parameters.
|
|
* Make the display of TCQ and queue depth clearer.
|
|
|
|
Wed Jan 17 23:30 2001 Gerard Roudier
|
|
* version sym-2.1.3-20010117
|
|
- Wrong residual values were returned in some situations.
|
|
This broke cdrecord with linux-2.4.0, for example.
|
|
|
|
Sat Jan 20 18:00 2001 Gerard Roudier
|
|
* version sym-2.1.4-20010120
|
|
- Add year 2001 to Copyright.
|
|
- A tiny bug in the dma memory freeing path has been fixed.
|
|
(Driver unload failed with a bad address reference).
|
|
|
|
Wed Jan 24 21:00 2001 Gerard Roudier
|
|
* version sym-2.1.5-20010124
|
|
- Make the driver work under Linux-2.4.x when statically linked
|
|
with the kernel.
|
|
- Check against memory allocation failure for SCRIPTZ and add the
|
|
missing free of this memory on instance detach.
|
|
- Check against GPIO3 pulled low for HVD controllers (driver did
|
|
just the opposite).
|
|
Misdetection of BUS mode was triggered on module reload only,
|
|
since BIOS settings were trusted instead on first load.
|
|
|
|
Wed Feb 7 21:00 2001 Gerard Roudier
|
|
* version sym-2.1.6-20010207
|
|
- Call pci_enable_device() as wished by kernel maintainers.
|
|
- Change the sym_queue_scsiio() interface.
|
|
This is intended to simplify portability.
|
|
- Move the code intended to deal with the dowloading of SCRIPTS
|
|
from SCRIPTS :) in the patch method (was wrongly placed in
|
|
the SCRIPTS setup method).
|
|
- Add a missing cpu_to_scr() (np->abort_tbl.addr)
|
|
- Remove a wrong cpu_to_scr() (np->targtbl_ba)
|
|
- Cleanup a bit the PPR failure recovery code.
|
|
|
|
Sat Mar 3 21:00 2001 Gerard Roudier
|
|
- Add option SYM_OPT_ANNOUNCE_TRANSFER_RATE and move the
|
|
corresponding code to file sym_misc.c.
|
|
Also move the code that sniffes INQUIRY to sym_misc.c.
|
|
This allows to share the corresponding code with NetBSD
|
|
without polluating the core driver source (sym_hipd.c).
|
|
- Add optionnal code that handles IO timeouts from the driver.
|
|
(not used under Linux, but required for NetBSD)
|
|
- Donnot assume any longer that PAGE_SHIFT and PAGE_SIZE are
|
|
defined at compile time, as at least NetBSD uses variables
|
|
in memory for that.
|
|
- Refine a work-around for the C1010-33 that consists in
|
|
disabling internal LOAD/STORE. Was applied up to revision 1.
|
|
Is now only applied to revision 0.
|
|
- Some code reorganisations due to code moves between files.
|
|
|
|
Tues Apr 10 21:00 2001 Gerard Roudier
|
|
* version sym-2.1.9-20010412
|
|
- Reset 53C896 and 53C1010 chip according to the manual.
|
|
(i.e.: set the ABRT bit in ISTAT if SCRIPTS are running)
|
|
- Set #LUN in request sense only if scsi version <= 2 and
|
|
#LUN <= 7.
|
|
- Set busy_itl in LCB to 1 if the LCB is allocated and a
|
|
SCSI command is active. This is a simplification.
|
|
- In sym_hcb_free(), do not scan the free_ccbq if no CCBs
|
|
has been allocated. This fixes a panic if attach failed.
|
|
- Add DT/ST (double/simple transition) in the transfer
|
|
negotiation announce.
|
|
- Forces the max number of tasks per LUN to at least 64.
|
|
- Use pci_set_dma_mask() for linux-2.4.3 and above.
|
|
- A couple of comments fixes.
|
|
|
|
Wed May 22:00 2001 Gerard Roudier
|
|
* version sym-2.1.10-20010509
|
|
- Mask GPCNTL against 0x1c (was 0xfc) for the reading of the NVRAM.
|
|
This ensure LEDC bit will not be set on 896 and later chips.
|
|
Fix sent by Chip Salzenberg <chip@perlsupport.com>.
|
|
- Define the number of PQS BUSes supported.
|
|
Fix sent by Stig Telfer <stig@api-networks.com>
|
|
- Miscellaneous common code rearrangements due to NetBSD accel
|
|
ioctl support, without impact on Linux (hopefully).
|
|
|
|
Mon July 2 12:00 2001 Gerard Roudier
|
|
* version sym-2.1.11-20010702
|
|
- Add Tekram 390 U2B/U2W SCSI LED handling.
|
|
Submitted by Chip Salzenberg <chip@valinux.com>
|
|
- Add call to scsi_set_pci_device() for kernels >= 2.4.4.
|
|
- Check pci dma mapping failures and complete the IO with some
|
|
error when such mapping fails.
|
|
- Fill in instance->max_cmd_len for kernels > 2.4.0.
|
|
- A couple of tiny fixes ...
|
|
|
|
Sun Sep 9 18:00 2001 Gerard Roudier
|
|
* version sym-2.1.12-20010909
|
|
- Change my email address.
|
|
- Add infrastructure for the forthcoming 64 bit DMA addressing support.
|
|
(Based on PCI 64 bit patch from David S. Miller)
|
|
- Donnot use anymore vm_offset_t type.
|
|
|
|
Sat Sep 15 20:00 2001 Gerard Roudier
|
|
* version sym-2.1.13-20010916
|
|
- Add support for 64 bit DMA addressing using segment registers.
|
|
16 registers for up to 4 GB x 16 -> 64 GB.
|
|
|
|
Sat Sep 22 12:00 2001 Gerard Roudier
|
|
* version sym-2.1.14-20010922
|
|
- Complete rewrite of the eh handling. The driver is now using a
|
|
semaphore in order to behave synchronously as required by the eh
|
|
threads. A timer is also used to prevent from waiting indefinitely.
|
|
|
|
Sun Sep 30 17:00 2001 Gerard Roudier
|
|
* version sym-2.1.15-20010930
|
|
- Include <linux/module.h> unconditionnaly as expected by latest
|
|
kernels.
|
|
- Use del_timer_sync() for recent kernels to kill the driver timer
|
|
on module release.
|
|
|
|
Sun Oct 28 15:00 2001 Gerard Roudier
|
|
* version sym-2.1.16-20011028
|
|
- Slightly simplify driver configuration.
|
|
- Prepare a new patch against linux-2.4.13.
|
|
|
|
Sat Nov 17 10:00 2001 Gerard Roudier
|
|
* version sym-2.1.17
|
|
- Fix a couple of gcc/gcc3 warnings.
|
|
- Allocate separately from the HCB the array for CCBs hashed by DSA.
|
|
All driver memory allocations are now not greater than 1 PAGE
|
|
even on PPC64 / 4KB PAGE surprising setup.
|
|
|
|
Sat Dec 01 18:00 2001 Gerard Roudier
|
|
* version sym-2.1.17a
|
|
- Use u_long instead of U32 for the IO base cookie. This is more
|
|
consistent with what archs are expecting.
|
|
- Use MMIO per default for Power PC instead of some fake normal IO,
|
|
as Paul Mackerras stated that MMIO works fine now on this arch.
|