On Tuesday 08 July 2008, Sergei Shtylyov wrote:
[...]
> >>>Index: b/drivers/ide/ide-iops.c
> >>>===================================================================
> >>>--- a/drivers/ide/ide-iops.c
> >>>+++ b/drivers/ide/ide-iops.c
> >>>@@ -688,8 +688,7 @@ int ide_driveid_update(ide_drive_t *driv
> >>> */
> >>>
> >>> SELECT_MASK(drive, 1);
> >>>- if (IDE_CONTROL_REG)
> >>>- hwif->OUTB(drive->ctl,IDE_CONTROL_REG);
> >>>+ ide_set_irq(drive, 1);
>
> >> If we're going to execute the command using polling, isn't it logical to
> >>*disable* drive's interrupt instead of enabling it which this code is
> >>currently doing? This looks like it might work only for the drivers having
> >>the maskproc() method (of which hpt366.c is the only one that I've ever dealt
> >>with).
>
> > Yes, this needs fixing.
>
> But should we honor drive->quirk_list here? What its different values
> mean? I'm seeing either 1 or 2 is used to decide whether to set nIEN or not...
I did some research on ->quirk_list in the past but I couldn't exactly
figure it out. I also wasn't able to trace this code to its author
(not that I tried very hard)...
Anyway my current findings/theories are the following:
- the diff between hpt366 and pdc202xx_{new,old} quirky devices:
--- hpt366.c 2008-04-12 22:03:54.000000000 +0200
+++ pdc202xx_new.c 2008-04-12 22:04:03.000000000 +0200
@@ -1,7 +1,11 @@
-static const char *quirk_drives[] = {
+static const char *pdc_quirk_drives[] = {
"QUANTUM FIREBALLlct08 08",
"QUANTUM FIREBALLP KA6.4",
+ "QUANTUM FIREBALLP KA9.1",
"QUANTUM FIREBALLP LM20.4",
+ "QUANTUM FIREBALLP KX13.6",
+ "QUANTUM FIREBALLP KX20.5",
+ "QUANTUM FIREBALLP KX27.3",
"QUANTUM FIREBALLP LM20.5",
NULL
};
indicates that pdc202xx_{new,old} lists lack few devices that were added
only to hpt366 and need fixing
- different values 1 (used by hpt366) or 2 (used by pdc202xx_new) are used
stricly for historical reasons (IOW they are the result of people hitting
problems with nIEN in different places and lack of generic solution) so
we should make ->quirk_list a single bit flag and enable all quirks for
both host drivers
- the problem most likely is not limited to hpt366/pdc202xx_{new,old}
so we should move its handling to core IDE code
- we should make sure that all SELECT_MASK() users also call ide_set_irq()
(then sgiioc4's ->maskproc can be removed since it is a copy of ->set_irq)
- we may want to make SELECT_MASK() call enable/disable_irq() for
->quirk_list device if there is no >maskproc method
- we need to be a bit careful with hpt366's ->maskproc because it checks
for ->quirk_list internally (so either we need to mask/unmask for all
devices on hpt366 or limit masking/unmasking to ->quirk_list on icside)
- we should merge SELECT_MASK() with ide_set_irq() and consider skipping
ide_set_irq() for ->quirk_list devices
There is an added bonus in fixing the above issues as it would result in
(finally) having sane API for IRQ (un)masking, namely:
* ->set_irq in struct ide_tp_ops for device's side
* ->maskproc in struct ide_port_ops for controller's side
As usual, I'll be glad to help with any patches.
Thanks,
Bart
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
