[lkml]   [2005]   [Mar]   [4]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
Subject[PATCH 2.6.11-rc3 00/08] ide: taskfile cleanup

Hello, Bartlomiej.
Hello, Jeff.

These eight patches

* define ATA_TFLAG_{OUT|IN}_* flags
* unify/generalize taskfile transport
* cleanup ide driver accordingly

For behavior changes by #03. I don't think defining a special flag
to handle the TASKFILE case is necessary. The change isn't
user-visible. And for the TASK ioctl, I think we should modify all
ioctls to disallow changing the upper nibble of the DEVICE register
except for the LBA bit. What do you think?

I tried hard not to break things and tested changes but I'm pretty
sure that I've missed something. So, please comment. :-)

[ Start of patch descriptions ]

: add individual ATA_TFLAG_{OUT|IN}_* flags

This patch replaces ide_task_t->tf_{out|in}_flags handling
with newly defined individual ATA_TFLAG_{OUT|IN}_* flags and
helper functions ide_{load|read}_taskfile(). To ease
transition of the IDE code, temporary flags
This patch is tit-for-tat and shouldn't change any behavior.

: convert __ide_do_rw_disk() to use ide_load_taskfile()

Reimplements __ide_do_rw_disk() using ide_load_taskfile().
While at it, clean up the function a little bit.

: remove flagged_taskfile() and unify taskfile paths

This patch removes flagged_taskfile(). All taskfile command
issuing goes through do_rw_taskfile(). do_rw_taskfile()
doesn't modify mangle with load flags anymore. It's now
caller's responsibility to set appropriate flags. Likewise,
ide_end_drive_cmd() is modified not to mangle with read flags,
and ide_dma_intr() now also finishes commands with
task_end_request(). Above changes make taskfile path unified
& generic.

As all ioctl subtleties are now responsibility of respective
ioctl functions. TASKFILE and TASK ioctl functions are
updated to set flags according to old behaviors. The
following two behavior changes occur.

* TASKFILE ioctl: taskfile registers are read back whether or
not the command fails. As copying back to user doesn't
happen in cases where reading back didn't occur before, this
change isn't user-visible. Defining & using a flag like
ATA_TFLAG_READ_ON_ERROR will remove this issue.
* TASK ioctl: drive->select.all & ~ATA_LBA is OR'd to device
value. Previously, only ATA_DEV bit was OR'd.

Also, all ide_{raw|diag}_taskfile(), do_rw_taskfile() users
are converted to use the new individual OUT/IN flags. As
results, the following behavior changes occur.

* idedisk_read_native_max_address(): ADDR/LBA48 regs are not
loaded. LBA48/DEVICE registers are not read back unless
* idedisk_set_max_address(): DEVICE register is not read
unless necessary.
* smart_enable(): DEVICE register is not loaded. Registers
are not read back.
* smart_disable(): ditto
* get_smart_threshold(): DEVICE register is not loaded.
* ide_task_init_flush(): ADDR/LBA48/DEVICE registers are not
* ide_init_specify_cmd(): Register aren't read back.
* ide_init_restore_cmd(): DEVICE register not loaded. No read back.
* ide_init_setmult_cmd(): ditto

: remove unused fields ide_drive_t->rq and ide_task_t->special

Remove unused fields ide_drive_t->rq and ide_task_t->special

: use ide_task_t->tf.protocol instead of ide_task_t->data_phase

Remove ide_task_t->{data_phase,command_type,prehandler,rq} and
use tf->protocol instead. Now the protocol value wholey
defines how to drive a taskfile except for NODATA cases where
a caller can optionally specify handler (for special
commands). The following behavior changes occur.

* ide_taskfile_ioctl(): req_task->command_type is ignored.
This doesn't make any difference except for error/crash
cases in the original code.

: convert set_xfer_rate() to use taskfile ioctl

Convert set_xfer_rate() to use taskfile ioctl.

: reimplement ide_cmd_ioctl() using taskfile

Reimplement ide_cmd_ioctl() using taskfile.

: remove REQ_DRIVE_CMD handling

Remove REQ_DRIVE_CMD handling. ide_init_drive_cmd() now
defaults to REQ_DRIVE_TASKFILE (now the only drive command :-).

[ End of patch descriptions ]


To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to
More majordomo info at
Please read the FAQ at

 \ /
  Last update: 2005-03-22 14:10    [W:0.056 / U:0.812 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site