USB broken in 2.4.0

USB broken in 2.4.0

[Heitzso]

I just tested with fresh-out-of-the-box
2.4.0 and using the newer libusb 0.1.2 
as suggested by antirez  (see email chain
below for more info).  I compiled libusb
and s10sh code this AM under 2.4.0. 

It blows up BAD by finding increasingly
larger photo images in the camera over
the usb link and extracting them to disk.
So by the third file you're trying to
extract gig sized files.  Obviously the
filesystem files up, the sytem chokes, etc.

This is the same code that works fine
under 2.2.18 kernel (I use it all of the
time there).


Heitzso

-----Original Message-----
From: antirez [mailto:antirez@invece.org]
Sent: Wednesday, January 03, 2001 6:00 PM
To: Heitzso
Cc: 'Johannes Erdfelt'; 'antirez@invece.org'
Subject: Re: usb broken in 2.4.0 prerelease versus 2.2.18


On Wed, Jan 03, 2001 at 12:33:34PM -0500, Heitzso wrote:

Hi Heitzso!

> (Salvatore Sanfilippo ... I'm cc'ing you
> to note that s10sh -u (usb mode) works fine
> with 2.2.18 but is breaking under 2.4.0 prerelease.

Too bad. The problem can't be of s10sh itself I guess.
It's more likely that some USB userspace ioctl() changed
so the USB subsystem isn't access without problems,
even if it's possible that some s10sh bug grow up
with a more robust USB implementation in 2.4.0.
I tested s10sh in 2.4.0-testxxx and it worked:
I suggest to try the new libusb 0.1.2 at
http://sourceforge.net/project/showfiles.php?group_id=1674&release_id=13538
and try to make s10sh working with this updated library.
I'll try. If you is able to fix s10sh or to make it working
using modified libusb version please drop me a mail!

thanks for your effort,
regars,
antirez

-- 
Salvatore Sanfilippo              |
<antirez@invece.org>
http://www.kyuzz.org/antirez      |      PGP: finger
antirez@tella.alicom.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: USB broken in 2.4.0

[Greg KH]

On Fri, Jan 05, 2001 at 12:38:25PM -0500, Heitzso wrote:
> I just tested with fresh-out-of-the-box
> 2.4.0 and using the newer libusb 0.1.2 
> as suggested by antirez  (see email chain
> below for more info).  I compiled libusb
> and s10sh code this AM under 2.4.0. 
> 
> It blows up BAD by finding increasingly
> larger photo images in the camera over
> the usb link and extracting them to disk.
> So by the third file you're trying to
> extract gig sized files.  Obviously the
> filesystem files up, the sytem chokes, etc.
> 
> This is the same code that works fine
> under 2.2.18 kernel (I use it all of the
> time there).

I made the same request to Jordan Mendelson yesterday, who has the same
problem.  Could you be so kind as to try to narrow down which kernel
version this broke on?  I have reports that it used to work on -test9
but doesn't now.  Could you try -test10, etc and let me know?

thanks,

greg k-h

-- 
greg@(kroah|wirex).com
http://immunix.org/~greg
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: USB broken in 2.4.0

[antirez]

On Fri, Jan 05, 2001 at 10:00:40AM -0800, Greg KH wrote:
> I made the same request to Jordan Mendelson yesterday, who has the same
> problem.  Could you be so kind as to try to narrow down which kernel
> version this broke on?  I have reports that it used to work on -test9
> but doesn't now.  Could you try -test10, etc and let me know?

I'll do some test with the new 2.4 kernel to find if there is a problem
in s10sh itself. A good test can be to try if the equivalent driver
of gphoto works without problems using the 2.4 kernel (however it also
uses the libusb). The s10sh bug may be not necessarly related to the USB
subsystem.

regards,
antirez

-- 
Salvatore Sanfilippo              |                      <antirez@invece.org>
http://www.kyuzz.org/antirez      |      PGP: finger antirez@tella.alicom.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

RE: USB broken in 2.4.0

[Dunlap, Randy]

This rings a small bell with me.
There was a change by Dan Streetman IIRC to limit
usbdevfs bulk transfers to PAGE_SIZE (4 KB for x86,
or 0x1000).  Anything larger than that returns
an error (-EINVAL).

~Randy
_______________________________________________
|randy.dunlap_at_intel.com        503-677-5408|
|NOTE: Any views presented here are mine alone|
|& may not represent the views of my employer.|
-----------------------------------------------

> -----Original Message-----
> From: antirez [mailto:antirez@invece.org]
> Sent: Friday, January 05, 2001 6:40 PM
> To: antirez
> Cc: Greg KH; Heitzso; 'linux-kernel@vger.kernel.org'; 
> 'Johannes Erdfelt'
> Subject: Re: USB broken in 2.4.0
> 
> 
> On Sat, Jan 06, 2001 at 12:04:29AM +0100, antirez wrote:
> > I'll do some test with the new 2.4 kernel to find if there 
> is a problem
> > in s10sh itself. A good test can be to try if the equivalent driver
> > of gphoto works without problems using the 2.4 kernel 
> (however it also
> > uses the libusb). The s10sh bug may be not necessarly 
> related to the USB
> > subsystem.
> 
> Ok, the problem is the same that I ecountered developing the file
> upload for the powershot USB driver performing a bulk write with
> a big data size, but this time it is present even reading.
> 
> s10sh reads 0x1400 bytes at once downloading jpges from the
> digicam, but the ioctl() that performs the bulk read fails with 2.4
> using this size. If I resize it (for example to 0x300) it 
> works without
> problems (with high performace penality, of course, 60% of slow-down).
> I don't know why. I checked at the time of the file upload the kernel
> code finding nothing.
> 
> Anyway with the old releases of the USB subsystem libusb failed to
> initialize the camera some time, now it seems fixed.
> 
> For the users: just edit usb.c and check the function USB_get_data(),
> substituting all the occurrence of 0x1400 to 0x300 as a work-around.
> 
> Please CC: me since I'm not subscribed to the list.
> 
> regards,
> antirez
> 
> -- 
> Salvatore Sanfilippo              |                      
> <antirez@invece.org>
> http://www.kyuzz.org/antirez      |      PGP: finger 
> antirez@tella.alicom.com
> -
> To unsubscribe from this list: send the line "unsubscribe 
> linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> Please read the FAQ at http://www.tux.org/lkml/
> 

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: USB broken in 2.4.0

[Johannes Erdfelt]

On Sat, Jan 06, 2001, antirez <antirez@invece.org> wrote:
> On Fri, Jan 05, 2001 at 04:48:00PM -0800, Dunlap, Randy wrote:
> > This rings a small bell with me.
> > There was a change by Dan Streetman IIRC to limit
> > usbdevfs bulk transfers to PAGE_SIZE (4 KB for x86,
> > or 0x1000).  Anything larger than that returns
> > an error (-EINVAL).
> 
> Yes, devio.c, proc_bulk():
> 
>         if (len1 > PAGE_SIZE)
>                 return -EINVAL;
> 
> Actually it is the max transfer size I can reach.
> I guess that to limit the page size can be an impementation
> advantage but it may slow-down a bit some userspace driver.
> I feel that even if the linux way is to implement the USB
> drivers in kernel space a full-featured USB user space access
> should be allowed.

This is the responsibility of the user space access. Using 4k reads
versus 8k reads won't increase performance (noticably), or will it
function any more correctly.

The next version of libusb will split up reads correctly so applications
using the library don't need to worry about this.

JE

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: USB broken in 2.4.0

[antirez]

On Sat, Jan 06, 2001 at 12:04:29AM +0100, antirez wrote:
> I'll do some test with the new 2.4 kernel to find if there is a problem
> in s10sh itself. A good test can be to try if the equivalent driver
> of gphoto works without problems using the 2.4 kernel (however it also
> uses the libusb). The s10sh bug may be not necessarly related to the USB
> subsystem.

Ok, the problem is the same that I ecountered developing the file
upload for the powershot USB driver performing a bulk write with
a big data size, but this time it is present even reading.

s10sh reads 0x1400 bytes at once downloading jpges from the
digicam, but the ioctl() that performs the bulk read fails with 2.4
using this size. If I resize it (for example to 0x300) it works without
problems (with high performace penality, of course, 60% of slow-down).
I don't know why. I checked at the time of the file upload the kernel
code finding nothing.

Anyway with the old releases of the USB subsystem libusb failed to
initialize the camera some time, now it seems fixed.

For the users: just edit usb.c and check the function USB_get_data(),
substituting all the occurrence of 0x1400 to 0x300 as a work-around.

Please CC: me since I'm not subscribed to the list.

regards,
antirez

-- 
Salvatore Sanfilippo              |                      <antirez@invece.org>
http://www.kyuzz.org/antirez      |      PGP: finger antirez@tella.alicom.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: USB broken in 2.4.0

[antirez]

On Sat, Jan 06, 2001 at 03:39:36AM +0100, antirez wrote:
> s10sh reads 0x1400 bytes at once downloading jpges from the
> digicam, but the ioctl() that performs the bulk read fails with 2.4
> using this size. If I resize it (for example to 0x300) it works without
> problems (with high performace penality, of course, 60% of slow-down).

I checked it better: the largest bulk read I can perform is of 4096 bytes
(up to 5120 bytes with old kernels).

antirez

(please CC me)

-- 
Salvatore Sanfilippo              |                      <antirez@invece.org>
http://www.kyuzz.org/antirez      |      PGP: finger antirez@tella.alicom.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: USB broken in 2.4.0

[antirez]

On Fri, Jan 05, 2001 at 04:48:00PM -0800, Dunlap, Randy wrote:
> This rings a small bell with me.
> There was a change by Dan Streetman IIRC to limit
> usbdevfs bulk transfers to PAGE_SIZE (4 KB for x86,
> or 0x1000).  Anything larger than that returns
> an error (-EINVAL).

Yes, devio.c, proc_bulk():

        if (len1 > PAGE_SIZE)
                return -EINVAL;

Actually it is the max transfer size I can reach.
I guess that to limit the page size can be an impementation
advantage but it may slow-down a bit some userspace driver.
I feel that even if the linux way is to implement the USB
drivers in kernel space a full-featured USB user space access
should be allowed.

antirez
(please cc me)

-- 
Salvatore Sanfilippo              |                      <antirez@invece.org>
http://www.kyuzz.org/antirez      |      PGP: finger antirez@tella.alicom.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

RE: USB broken in 2.4.0

[Heitzso]

There's a flurry of emails I haven't read yet.
I wanted to answer the direction question first

 test9 OK
 test10 OK
 test11 BROKEN

so that's where the break occurred.

I've gotta admit I was scared running the older
test kernels because somewhere in the series
the filesystem could be damaged and it bit me
bad on the first pass a couple of months ago.
I compiled the kernels under current full 2.4.0
and when rebooting to the test kernels tried
to get in, do the test, and get out as fast
as possible.

Heitzso


-----Original Message-----
From: Greg KH [mailto:greg@wirex.com]
Sent: Friday, January 05, 2001 1:01 PM
To: Heitzso
Cc: 'antirez@invece.org'; 'linux-kernel@vger.kernel.org'; 'Johannes
Erdfelt'
Subject: Re: USB broken in 2.4.0


On Fri, Jan 05, 2001 at 12:38:25PM -0500, Heitzso wrote:
> I just tested with fresh-out-of-the-box
> 2.4.0 and using the newer libusb 0.1.2 
> as suggested by antirez  (see email chain
> below for more info).  I compiled libusb
> and s10sh code this AM under 2.4.0. 
> 
> It blows up BAD by finding increasingly
> larger photo images in the camera over
> the usb link and extracting them to disk.
> So by the third file you're trying to
> extract gig sized files.  Obviously the
> filesystem files up, the sytem chokes, etc.
> 
> This is the same code that works fine
> under 2.2.18 kernel (I use it all of the
> time there).

I made the same request to Jordan Mendelson yesterday, who has the same
problem.  Could you be so kind as to try to narrow down which kernel
version this broke on?  I have reports that it used to work on -test9
but doesn't now.  Could you try -test10, etc and let me know?

thanks,

greg k-h

-- 
greg@(kroah|wirex).com
http://immunix.org/~greg
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/

Re: USB broken in 2.4.0

[antirez]

On Mon, Jan 08, 2001 at 10:38:31AM -0500, Heitzso wrote:
> so that's where the break occurred.

The problem was fixed (new interface don't allow
a bulk read to be more than PAGE_SIZE, often 4096 bytes)
Read the thread for more information.
You can download the fixed s10sh at
http://www.kyuzz.org/antirez/s10sh.html

antirez

-- 
Salvatore Sanfilippo              |                      <antirez@invece.org>
http://www.kyuzz.org/antirez      |      PGP: finger antirez@tella.alicom.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/