Andy Kellner (from ConnectCom Solutions formerly
known as Advansys) and Bob Frey (former maintainer)
working in concert have posted several "3.3x" versions
of the advansys driver to the linux-scsi list. Despite
this, there seems no sign of this improved driver
being included in the 2.4 series kernels. However
the advansys driver was upgraded to version 3.3D in
lk 2.2.18 . Have there been any adverse reports about
the advansys driver in lk 2.2.18 ?
I am currently using advansys driver version 3.3G
without problems on lk 2.4.2 . The changelog at the
top of the advansys.c file shows an impressive number
of improvements and fixes between versions 3.2M and
3.3G (including powerpc support, see below).
This is not the first time that I've sent such a post
trying to press for an update of this driver. I was
told that the patch was too big to be contemplated for
lk 2.4 . Well size didn't stop the complete replacement
of the much used aic7xxx driver.
You do have some short term options. Since the
advansys driver has the same source in the lk 2.2
and lk 2.4 series, you can copy the version 3.3D
advansys.[hc] files from your lk 2.2.18 source to
your lk 2.4.2 source and it will build ok.
Alternatively you can get a recent version (version
3.3F) from:
http://www.connectcom.net/support/evaluation.html
Doug Gilbert
Changelog from advansys.c file between versions
3.2M and 3.3G follows:
vvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvvv
3.2N (4/1/00):
1. Add CONFIG_ISA ifdef code.
2. Include advansys_interrupts_enabled name change patch.
3. For >= v2.3.28 use new SCSI error handling with new function
advansys_eh_bus_reset(). Don't include an abort function
because of base library limitations.
4. For >= v2.3.28 use per board lock instead of io_request_lock.
5. For >= v2.3.28 eliminate advansys_command() and
advansys_command_done().
6. Add some changes for PowerPC (Big Endian) support, but it isn't
working yet.
7. Fix "nonexistent resource free" problem that occurred on a module
unload for boards with an I/O space >= 255. The 'n_io_port' field
is only one byte and can not be used to hold an ioport length more
than 255.
3.3A (4/4/00):
1. Update to Adv Library 5.8.
2. For wide cards add support for CDBs up to 16 bytes.
3. Eliminate warnings when CONFIG_PROC_FS is not defined.
3.3B (5/1/00):
1. Support for PowerPC (Big Endian) wide cards. Narrow cards
still need work.
2. Change bitfields to shift and mask access for endian
portability.
3.3C (10/13/00):
1. Update for latest 2.4 kernel.
2. Test ABP-480 CardBus support in 2.4 kernel - works!
3. Update to Asc Library S123.
4. Update to Adv Library 5.12.
3.3D (11/22/00):
1. Update for latest 2.4 kernel.
2. Create patches for 2.2 and 2.4 kernels.
3.3E (1/9/01):
1. Now that 2.4 is released remove ifdef code for kernel versions
less than 2.2. The driver is now only supported in kernels 2.2,
2.4, and greater.
2. Add code to release and acquire the io_request_lock in
the driver entrypoint functions: advansys_detect and
advansys_queuecommand. In kernel 2.4 the SCSI mid-level driver
still holds the io_request_lock on entry to SCSI low-level drivers.
This was supposed to be removed before 2.4 was released but never
happened. When the mid-level SCSI driver is changed all references
to the io_request_lock should be removed from the driver.
3. Simplify error handling by removing advansys_abort(),
AscAbortSRB(), AscResetDevice(). SCSI bus reset requests are
now handled by resetting the SCSI bus and fully re-initializing
the chip. This simple method of error recovery has proven to work
most reliably after attempts at different methods. Also now only
support the "new" error handling method and remove the obsolete
error handling interface.
4. Fix debug build errors.
3.3F (1/24/01):
1. Merge with ConnectCom version from Andy Kellner which
updates Adv Library to 5.14.
2. Make PowerPC (Big Endian) work for narrow cards and
fix problems writing EEPROM for wide cards.
3. Remove interrupts_enabled assertion function.
3.3G (2/16/01):
1. Return an error from narrow boards if passed a 16 byte
CDB. The wide board can already handle 16 byte CDBs.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/