Re: Bug in 2.6.29 ide-cd: Kernel freeze: bisected + unacceptable workaround

[Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>]

On Monday 30 March 2009, Borislav Petkov wrote:
> Hi Michael,
> 
> On Sun, Mar 29, 2009 at 11:37:40PM +0200, Michael Roth wrote:
> > Hello Borislav,
> > 
> > thank you for your help.
> > 
> > Borislav Petkov schrieb:
> > 
> > >  Can we please see a dmesg of a working
> > > kernel on your machine
> > 
> > Sure, dmesg output appended beneath. I also added the dmesg output
> > of the patched kernel. See below.
> > 
> > 
> > > and also is it possible to catch any kernel output of the crash
> > > any oops or something when your machine freezes - the more the better.
> > 
> > There aren't any messages, neither on the console nor in the syslog.
> > The kernel simple freeze, cursor on the console stops blinking, no
> > more ide activity. I have to hit the reset-bottom to restart.
> > 
> > 
> > > Probably over a serial console would be best but photographed
> > > with a digicam and uploaded somewhere as a screenshot would do too.
> > 
> > I will do that on request, but it will take some time (about days)
> > because currently I don't have access to a second computer or so.
> 
> That'll definitely be very helpful since I wanna know exactly why it
> locks up. If you still have no output it'd be cool if you could enable
> ide-cd debugging and retry catching the output. Here's a patch for that,
> just recompile and reboot (I admit this is not one of the finest ways to
> enable debugging but since this is work in progress...:) ):
> 
> --
> diff --git a/drivers/ide/ide-atapi.c b/drivers/ide/ide-atapi.c
> index e9d042d..4ef3b7e 100644
> --- a/drivers/ide/ide-atapi.c
> +++ b/drivers/ide/ide-atapi.c
> @@ -8,6 +8,8 @@
>  #include <linux/ide.h>
>  #include <scsi/scsi.h>
>  
> +#define DEBUG 1
> +
>  #ifdef DEBUG
>  #define debug_log(fmt, args...) \
>  	printk(KERN_INFO "ide: " fmt, ## args)
> diff --git a/drivers/ide/ide-cd.c b/drivers/ide/ide-cd.c
> index ddfbea4..ea36295 100644
> --- a/drivers/ide/ide-cd.c
> +++ b/drivers/ide/ide-cd.c
> @@ -1991,7 +1991,7 @@ static int ide_cd_probe(ide_drive_t *drive)
>  		}
>  	}
>  
> -	drive->debug_mask = debug_mask;
> +	drive->debug_mask = 0xffffffff;
>  	drive->irq_handler = cdrom_newpc_intr;
>  
>  	info = kzalloc(sizeof(struct cdrom_info), GFP_KERNEL);
> diff --git a/drivers/ide/ide-cd.h b/drivers/ide/ide-cd.h
> index c878bfc..225afca 100644
> --- a/drivers/ide/ide-cd.h
> +++ b/drivers/ide/ide-cd.h
> @@ -8,7 +8,7 @@
>  #include <linux/cdrom.h>
>  #include <asm/byteorder.h>
>  
> -#define IDECD_DEBUG_LOG		0
> +#define IDECD_DEBUG_LOG		1
>  
>  #if IDECD_DEBUG_LOG
>  #define ide_debug_log(lvl, fmt, args...) __ide_debug_log(lvl, fmt, args)
> 
> > > You could also try the following patch to see whether that is of any
> > > help:
> > 
> > Your patch helps perfectly. I'm writing this email on the patched
> > 2.6.29. I also tested CD and DVD access: They work well. Burning not
> > tested but I assume it will work too.
> 
> @Bart: can you take at look at this. Somehow, if the device is behind a
> PCI IDE controller, the order of issuing the command and enabling DMA is
> important. Seen something like that before?

Spec is unclear on the ordering but empirically it seems that some hosts
may need packet command to start DMA.  I think that we should proceed with
your patch (please repost with patch description) and also apply the same
change to the rest of ATAPI devices in a subsequent patch.

Thanks,
Bart