summaryrefslogtreecommitdiffstats
path: root/sys/dev/pci/ips.c (follow)
Commit message (Collapse)AuthorAgeFilesLines
...
* When calculating physical disk size use its number of sectorsgrange2009-03-201-2/+2
| | | | rather than chunk's.
* Detect unused and hot-spare drives.grange2009-03-201-12/+44
|
* Support bio ``unused'' drives.grange2009-03-201-1/+2
|
* Cleanup physical drive states handling.grange2009-03-201-12/+15
|
* Fetch rebuild status only for degraded or offline volumes.grange2009-03-201-7/+16
|
* For rebuilds use a special REBUILD command instead of a genericgrange2009-03-191-3/+31
| | | | SETSTATE, for some reason it works better on lpinto's machine.
* For pass-through commands adjust timeout value to what controllergrange2009-03-191-4/+14
| | | | | supports. Make sure our timeout will be fired after controller gives up.
* Unify command id printing.grange2009-03-191-4/+5
|
* Make ips_timeout to be usable not only with scsi xfers.grange2009-03-191-13/+13
|
* Remove unused code.grange2009-03-191-65/+1
|
* Keep softc pointer in ccb because it requires some work to correctlygrange2009-03-181-2/+4
| | | | | extract it from scsi_xfer in ips_timeout since link->adapter_softc may point to a per-channel structure in pass-through path.
* Provide both errno.h codes for the userspace tools like bioctlgrange2009-03-171-21/+52
| | | | and XS_* codes for the scsi layer.
* Fix compilation without bio(4).grange2009-03-171-1/+5
|
* Implement bioctl volume management: rebuild and hotspare.grange2009-03-161-5/+109
|
* When doing pass-through calculate right xs->resid value based ongrange2009-03-161-4/+7
| | | | the info hardware returnes.
* Don't set error on data underrun. This tiny change allowed scsi tapegrange2009-03-161-1/+7
| | | | on the pass-through bus on lpinto's machine to work.
* Remove redundant DPRINTF from ips_morpheus_isintr.grange2009-03-161-7/+3
|
* IPS_SCSI_PT flag not used any more.grange2009-03-161-3/+1
|
* Cleanup error processing:grange2009-03-151-102/+96
| | | | | | | | - in ips_start_xs don't do scsi_done twice on error; - make ips_error return scsi XS_* codes rather than errno.h E* codes; - move all error reporting under debug, let upper layers to spam user. And fix a few missing byte conversions while here.
* Add SCSI pass-through and enclosure support. Tested on lpinto'sgrange2009-03-131-3/+214
| | | | machine which has safte(4).
* Some preparation steps for SCSI pass-through support:grange2009-03-131-32/+64
| | | | | | | | | | - stack command and DCDB frames along with scatter-gather list into one command block; - use structure field pointers and offsetof() instead of evil pointer arithmetics; - add DCDB definitions. No functional changes.
* Process command completion errors while polling, too.grange2009-03-131-18/+22
|
* In the most common case -- array failure -- move error message undergrange2009-03-121-1/+7
| | | | | debug not to spam dmesg, bioctl and sysctl hw.sensors will tell you enough.
* Advanced error checking.grange2009-03-121-21/+78
|
* For SYNCHRONIZE_CACHE command use the same code path as for READ/WRITE.grange2009-03-121-65/+61
| | | | This required to factor out some common scsi code.
* Rework command polling to be able to sleep if no SCSI_NOSLEEP flaggrange2009-03-111-134/+156
| | | | | passed. Sleep while waiting for sensors update because that's what sensors work queue for.
* Print scsi xfer flags in debug output.grange2009-03-101-3/+3
|
* Rearrange physical drive state bits checking so that bioctl getsgrange2009-03-101-5/+5
| | | | a correct state value.
* Add some more debug.grange2009-03-101-2/+25
|
* Add ips_intrds() function to be able to disable interrupts on hardware.grange2009-03-101-1/+23
|
* Limit registers space size to some reasonable value.grange2009-03-101-2/+4
|
* Completely rework command processing:grange2009-03-091-218/+342
| | | | | | | | | | | - use separate callbacks to finish scsi and raid management commands; - for raid management commands use dedicated dma-safe buffer; - remove unused run queue; The main goal of these changes is to fix a bug showing up on the lpinto's machine where READCONF command fails because it needs more than 1-page buffer and the buffer was not contiguos and controller doesn't support scatter-gather for raid management commands.
* Time to turn off debug by default. It shrinks ramdisk as a bonus.grange2009-03-011-3/+1
|
* Add sensors support.grange2009-03-011-1/+71
|
* - add missing letoh32grange2009-03-011-3/+3
| | | | - tweak product string
* Don't use scsi_xfer if we don't run a command came from the scsi layer.grange2009-02-191-4/+9
|
* - add more bioctl support, now physical drives info availablegrange2009-02-171-54/+164
| | | | - shrink dmesg strings
* Extend the scsi_adapter minphys() callback to take a struct scsi_link *miod2009-02-161-2/+2
| | | | | | as additional argument. This will allow intermediate layers between scsi devices such as sd and scsi host adapters to take appropriate action if necessary.
* When faking scsi commands never copy more than xs->datalen bytes.grange2009-02-131-21/+21
|
* Prettier fake scsi inquiry ascii strings.grange2009-02-111-4/+4
|
* Add a few missing byteorder conversion calls, tho I don't reallygrange2009-02-111-6/+6
| | | | | hope ips is usable on anything beside amd64 or i386 because of its x86 BIOS.
* Add missing copperhead chipset support bits.grange2009-02-101-15/+67
|
* Update copyright year.grange2009-02-101-2/+2
|
* Improve adapter type displaying.grange2009-02-091-15/+63
|
* Start on bioctl support for ips. Only adapter and volume info for now.grange2009-02-091-2/+111
|
* Protect ips_poll call with splbio.grange2009-02-091-2/+4
|
* Lower debug level back, looks like i've found that data corruption bug.grange2009-02-081-2/+2
|
* Put back bzero'ing cmd structure before filling it in which was lostgrange2009-02-081-1/+2
| | | | during driver rewritting at h2k7.
* Temporary enable full debug to make it easier to commit diffs.grange2009-02-081-2/+2
|
* ips is not a real scsi controller and not supposed to talk directly togrange2009-02-081-2/+4
| | | | | various flacky scsi hardware (at least until passthrough implemented). So reduce huge default scsibus timeout values to somewhat reasonable.
Copyright © 1996 – 2023 Jason A. Donenfeld. All Rights Reverse Engineered.