public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* 'False' IO error when copying from cdrom drive
@ 2006-11-23 19:48 Mathieu Fluhr
  2006-11-23 20:18 ` Sergey Vlasov
  0 siblings, 1 reply; 3+ messages in thread
From: Mathieu Fluhr @ 2006-11-23 19:48 UTC (permalink / raw)
  To: LKML

Dear all,

I am currently facing a really weird thing. It seems that it is related
to the block driver (or whatever else related to block devices).

I explain: First take this really simple program:
----8<------------------------------------------------------------------
#include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>


int main(int argc, char **argv)
{
  if(argc < 2)
    return 0;
  
  int iFD = open(argv[1], O_RDONLY | O_NONBLOCK);
  if(iFD == -1)
    perror("open");

  while(1)
    sleep(1);

  return 0;
}
----8<-----------------------------------------------------------------

Apparently, it just open a file in read only mode, and non-blocking. 
If you use it for opening an IDE DVD recorder, it will just grab a file
descriptor, without checking if the media is present or not.
 -> This is normally how the /dev/hdXX device file should be open to 
    use CDROM_SEND_PACKET ioctl.

(Note: Feel free to correct me if I am wrong)


Ok. Now take a full DVD (I tested DVD+R, +RW and -R), with more than 
4 300 000 000 bytes (very important :), and perform the following:

0. Open the tray of your recorder
1. Launch this small program above, passing the recorder device file as 
   argument and let it run in background.
2. then put the disc in the device and mount it
3. try to copy to whole content on the hard drive

-> You will get an error like the following:
  > kernel: attempt to access beyond end of device
  > kernel: hdb: rw=0, want=8388612, limit=8388604

If you do not launch this small program in background, everything works
like a charm.

What is really astonishing, is that no matter the media, no matter the
device, this 'limit' is ALWAYS 8388604. Which - as far as I debugged -
make exactly 4 294 965 248 bytes (that is where the 4 300 000 000 bytes
come from).

So am I think that there must be somehow a value corruption, as:
1. the limit shown in the kernel error log is wrong
2. whitout running this small prog, everything works fine.

Regards,
Mathieu


^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: 'False' IO error when copying from cdrom drive
  2006-11-23 19:48 'False' IO error when copying from cdrom drive Mathieu Fluhr
@ 2006-11-23 20:18 ` Sergey Vlasov
  2006-11-24 10:14   ` Mathieu Fluhr
  0 siblings, 1 reply; 3+ messages in thread
From: Sergey Vlasov @ 2006-11-23 20:18 UTC (permalink / raw)
  To: Mathieu Fluhr; +Cc: linux-kernel

[-- Attachment #1: Type: text/plain, Size: 1976 bytes --]

On Thu, 23 Nov 2006 20:48:25 +0100 Mathieu Fluhr wrote:

> I explain: First take this really simple program:
> ----8<------------------------------------------------------------------
> #include <sys/types.h>
> #include <sys/stat.h>
> #include <fcntl.h>
>
>
> int main(int argc, char **argv)
> {
>   if(argc < 2)
>     return 0;
>
>   int iFD = open(argv[1], O_RDONLY | O_NONBLOCK);
>   if(iFD == -1)
>     perror("open");
>
>   while(1)
>     sleep(1);
>
>   return 0;
> }
> ----8<-----------------------------------------------------------------
[..]
> Ok. Now take a full DVD (I tested DVD+R, +RW and -R), with more than
> 4 300 000 000 bytes (very important :), and perform the following:
>
> 0. Open the tray of your recorder
> 1. Launch this small program above, passing the recorder device file as
>    argument and let it run in background.

At this time the following code in drivers/ide/ide-cd.c:cdrom_read_toc()
is executed:

	/* Try to get the total cdrom capacity and sector size. */
	stat = cdrom_read_capacity(drive, &toc->capacity, &sectors_per_frame,
				   sense);
	if (stat)
		toc->capacity = 0x1fffff;

	set_capacity(info->disk, toc->capacity * sectors_per_frame);

Obviously, with the tray open cdrom_read_capacity() won't succeed,
therefore toc->capacity will be set to 0x1fffff.

> 2. then put the disc in the device and mount it
> 3. try to copy to whole content on the hard drive
>
> -> You will get an error like the following:
>   > kernel: attempt to access beyond end of device
>   > kernel: hdb: rw=0, want=8388612, limit=8388604

0x1fffff * 2048/512 == 8388604

Your background program is keeping the device open, which prevents
revalidation of capacity and other media information on subsequent
device opens.

When HAL polls CD-ROM devices to detect newly inserted media, it does
not keep the device open forever - instead, it opens and closes the
device every time.  Your program should either do the same thing, or
just depend on HAL events.

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

^ permalink raw reply	[flat|nested] 3+ messages in thread

* Re: 'False' IO error when copying from cdrom drive
  2006-11-23 20:18 ` Sergey Vlasov
@ 2006-11-24 10:14   ` Mathieu Fluhr
  0 siblings, 0 replies; 3+ messages in thread
From: Mathieu Fluhr @ 2006-11-24 10:14 UTC (permalink / raw)
  To: Sergey Vlasov; +Cc: linux-kernel

Okay, I understand the point. So this 0x1fffff is somehow a default
value right? (Humm why 0x1fffff by the way?)

But:
- when I open /dev/hdxx with the O_NONBLOCK flag, this is specially to
get a 'raw' access on the device. If I do not pass this flag, the open
syscall will fail if no media is in the device.

- when that mount command is used, the mount will fail if no medium is
inside. So wouldn't it be a good idea to recheck the toc at this time
(something like calling the .revalidate_disk function I mean) 

These are just suggestions... I am just really astonished that a single
user program can disturb the mount command with optical disc devices :D

Regards,
Mathieu

On Thu, 2006-11-23 at 23:18 +0300, Sergey Vlasov wrote:
> On Thu, 23 Nov 2006 20:48:25 +0100 Mathieu Fluhr wrote:
> 
> > I explain: First take this really simple program:
> > ----8<------------------------------------------------------------------
> > #include <sys/types.h>
> > #include <sys/stat.h>
> > #include <fcntl.h>
> >
> >
> > int main(int argc, char **argv)
> > {
> >   if(argc < 2)
> >     return 0;
> >
> >   int iFD = open(argv[1], O_RDONLY | O_NONBLOCK);
> >   if(iFD == -1)
> >     perror("open");
> >
> >   while(1)
> >     sleep(1);
> >
> >   return 0;
> > }
> > ----8<-----------------------------------------------------------------
> [..]
> > Ok. Now take a full DVD (I tested DVD+R, +RW and -R), with more than
> > 4 300 000 000 bytes (very important :), and perform the following:
> >
> > 0. Open the tray of your recorder
> > 1. Launch this small program above, passing the recorder device file as
> >    argument and let it run in background.
> 
> At this time the following code in drivers/ide/ide-cd.c:cdrom_read_toc()
> is executed:
> 
> 	/* Try to get the total cdrom capacity and sector size. */
> 	stat = cdrom_read_capacity(drive, &toc->capacity, &sectors_per_frame,
> 				   sense);
> 	if (stat)
> 		toc->capacity = 0x1fffff;
> 
> 	set_capacity(info->disk, toc->capacity * sectors_per_frame);
> 
> Obviously, with the tray open cdrom_read_capacity() won't succeed,
> therefore toc->capacity will be set to 0x1fffff.
> 
> > 2. then put the disc in the device and mount it
> > 3. try to copy to whole content on the hard drive
> >
> > -> You will get an error like the following:
> >   > kernel: attempt to access beyond end of device
> >   > kernel: hdb: rw=0, want=8388612, limit=8388604
> 
> 0x1fffff * 2048/512 == 8388604
> 
> Your background program is keeping the device open, which prevents
> revalidation of capacity and other media information on subsequent
> device opens.
> 
> When HAL polls CD-ROM devices to detect newly inserted media, it does
> not keep the device open forever - instead, it opens and closes the
> device every time.  Your program should either do the same thing, or
> just depend on HAL events.


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2006-11-24 10:16 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-11-23 19:48 'False' IO error when copying from cdrom drive Mathieu Fluhr
2006-11-23 20:18 ` Sergey Vlasov
2006-11-24 10:14   ` Mathieu Fluhr

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox