Re: Oddness in ahcisata driver

[Yorick Hardy <yhardy%uj.ac.za@localhost>]

I am afraid it did not work for me.

I generally see the following:

...
ahcisata0 at pci0 dev 31 function 2: Intel 82801HBM SATA Controller
ahcisata0: interrupting at ioapic0 pin 19 (irq 10)
ahcisata0: AHCI revision 1.1, 3 ports, 32 command slots, features 0xe720e080
atabus2 at ahcisata0 channel 0
atabus3 at ahcisata0 channel 1
atabus4 at ahcisata0 channel 2
...
ahcisata0 port 0: device present, speed: 1.5Gb/s
...
wd0 at atabus2 drive 0ahcisata0 port 1: PHY offline
...

Is it possible that different channels are interfering
with each other?

I am using the attached patch currently, which is still
not a correct fix.

Manuel Bouyer wrote:
> On Thu, Apr 17, 2008 at 12:08:17PM -0500, M Graff wrote:
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > Yorick Hardy wrote:
> > | This is not a correct fix, but would you be interested in looking here?
> > |
> > |   http://mail-index.netbsd.org/port-i386/2008/04/14/msg000371.html
> >
> > This patch suggests retrying after a short delay.  I've tested it and
> > while it does work, it does not ALWAYS work -- one time in 15 it has
> > failed me if I rapidly reboot.
> >
> > I suppose we might need a:
> >
> > ~  maxtries = 5;
> > ~  while (err == CMD_ERR && maxtries-- > 0) {
> > ~    delay(...);
> > ~    err = wd->atabus->...
> > ~  }
>
>
> Could you try with ahcisata_core.c rev 1.15 that I just commited ?
> It may help for this problem.


--
Kind regards,

Yorick Hardy

--- sys/dev/ata/wd.c.orig       2008-05-01 22:32:25.000000000 +0200
+++ sys/dev/ata/wd.c    2008-05-02 00:34:58.000000000 +0200
@@ -1774,8 +1774,14 @@
 int
 wd_get_params(struct wd_softc *wd, u_int8_t flags, struct ataparams *params)
 {
+       int err = wd->atabus->ata_get_params(wd->drvp, flags, params);
 
-       switch (wd->atabus->ata_get_params(wd->drvp, flags, params)) {
+       while (err == CMD_ERR && wd->drvp->state != READY) {
+               delay(100);
+               err = wd->atabus->ata_get_params(wd->drvp, flags, params);
+       }
+
+       switch (err) {
        case CMD_AGAIN:
                return 1;
        case CMD_ERR: