-Original Message-
From: David Brownell
To: Christopher Li
Cc: Oliver Neukum; '[EMAIL PROTECTED]'
Sent: 2/5/03 8:59 PM
Subject: Re: [linux-usb-devel] usbdevfs setup URB buffer size limit
No ...
>Toggle on control endpoints must reset at the beginning of
>each transaction
Agree.
>and
-info = usb_pipeout (urb->pipe)?
- TD_CC | TD_R | TD_DP_OUT | TD_T_DATA1 : TD_CC | TD_R | TD_DP_IN | TD_T_DATA1;
going to
+ info = usb_pipeout (urb->pipe)? (TD_CC | TD_R | TD_DP_OUT | TD_T_TOGGLE)
+ : (TD_CC | TD_R | TD_DP_IN | TD_T_TOGGLE);
No ...
Toggle on control en
Oliver Neukum wrote:
Hi,
the fix is obvious.
Right ... not a 2.5 issue, this is for "automagic resubmit".
Thanks for the fix ... Greg, please merge.
- Dave
diff -Nru a/drivers/usb/hcd/ehci-q.c b/drivers/usb/hcd/ehci-q.c
--- a/drivers/usb/hcd/ehci-q.c Thu Feb 6 01:16:01 2003
+++ b/drivers/u
Christopher Hoover wrote:
Perhaps it would be better to hand parse the structure out?
There's already kernel code that has to parse it out, for translating
32bit usbfs ioctls to 64bit ones ... something to consider.
That's certainly a workable solution. Any structures passed up from
user spa
John Homppi wrote:
Long description follows. the code to review is at the end of the message.
===
I have a device that uses interrupt OUT endpoints and I need a Linux driver.
It seems to me that the existing USB driver stack does not support OUTPUT on
interrupt end-points.
Actually
Duncan Sands wrote:
On Tuesday 04 February 2003 22:53, David Brownell wrote:
Duncan Sands wrote:
(1) is there any speed advantage to submitting multiple urbs to
a bulk endpoint, as compared to using a single urb which is
resubmitted in its completion handler?
Yes, most noticeably at high spe
FORTUNE LOTTO SOUTH AFRICA,
10 SUNSET AVE. ROODEPORT, 4412
JOHANNESBURG, SOUTH AFRICA.
TEL: 27 824255626 FAX: 27 115075911
FROM: THE DESK OF THE PROMOTIONS MANAGER,
INTERNATIONAL PROMOTIONS/PRIZE AWARD DEPARTMENT,
REF: FLS/5181/7390007/02
BAT
FORTUNE LOTTO SOUTH AFRICA,
10 SUNSET AVE. ROODEPORT, 4412
JOHANNESBURG, SOUTH AFRICA.
TEL: 27 824255626 FAX: 27 115075911
FROM: THE DESK OF THE PROMOTIONS MANAGER,
INTERNATIONAL PROMOTIONS/PRIZE AWARD DEPARTMENT,
REF: FLS/5181/7390007/02
BAT
> From: Christopher Li <[EMAIL PROTECTED]>
> Date: Wed, 5 Feb 2003 07:26:45 -0800
> Any comment?
> /* min 8 byte setup packet, max arbitrary */
> - if (uurb.buffer_length < 8 || uurb.buffer_length > PAGE_SIZE)
> + if (uurb.buffer_length < 8 || (uurb.buffer_le
John:
Thank you very much for your information. Our device is also based on the
Microchip PIC16C7XX-series USB-PIC.
And indeed, it seems that Microsoft's interrupt OUT endpoint support is a
little easier to work with, since the drivers for the device can be written
completely in user-space.
I
I test my patch on an ohci box and find a bug. The last status
stage is always use DATA1?
This patch seems works so far. The previous one is broken.
Any comment are welcome.
Thanks
Chris
= devio.c 1.9 vs edited =
--- 1.9/drivers/usb/devio.c Thu Apr 18 05:34:31 2002
+++ edited/devi
Hi,
the fix is obvious.
Regards
Oliver
You can import this changeset into BK by piping this whole message to:
'| bk receive [path to repository]' or apply the patch as usual.
===
[EMAIL PROTECTED], 2003-02
I am sorry forget to send to the list.
Here is the patch version 3.
I try to fix the setup buffer size for ohci according the ohci spec.
I might be wrong about it, I don't know much about ohci and I am yet
to find an ohci box to test it.
Any comment?
thanks
Chris
= devio.c 1.9 vs edited ==
> /* min 8 byte setup packet, max arbitrary */
> - if (uurb.buffer_length < 8 || uurb.buffer_length > PAGE_SIZE)
> + if (uurb.buffer_length < 8 || uurb.buffer_length > 64*1024)
Still one byte off. It must fit into 16 bits.
Regards
Oli
Hi,
The Stanford Checker discovered a memory leak in cdc-acm.
This patch to 2.5.59 fixes it. Please apply.
BTW, Vojtech, there was a possible memory leak identified
in usb/input/pid.c, but I wasn't sure about it.
~Randy
diff -Naur ./drivers/usb/class/cdc-acm.c%LEAK ./drivers/usb/class/cdc-acm.c
Hi,
I'm trying to get the Ego Technology MD100 to work under linux. It's
small MP3 player gadget which has 128MB of flash and a USB interface. It
identifies itself as a USB storage device. When I try to read from it
under linux, the process doing the read just hangs. The device works
under windows
From usb-ohci.c:
case PIPE_CONTROL: /* 1 TD for setup, 1 for ACK and 1 for every 4096 B */
size = (urb->transfer_buffer_length == 0)? 2:
(urb->transfer_buffer_length - 1) / 4096 + 3;
break;
Then it will allocate "size" number of TDs.
Do I miss anything?
Look at td_submit_urb() where it o
I try it again.
I also check the ohci driver. I think the current ohci driver
should be able to handle >4K buffers.
>From usb-ohci.c:
case PIPE_CONTROL: /* 1 TD for setup, 1 for ACK and 1 for every 4096 B */
size = (urb->transfer_buffer_length == 0)? 2:
(
> -Original Message-
> From: David Brownell [mailto:[EMAIL PROTECTED]]
> Sent: Tuesday, February 04, 2003 8:14 PM
> To: Christopher Li
> Cc: '[EMAIL PROTECTED]'
> Subject: Re: [linux-usb-devel] usbdevfs setup URB buffer size limit
>
>
> Christopher Li wrote:
> >
> > So usbde
On Wed, 5 Feb 2003 [EMAIL PROTECTED] wrote:
> Hello Alan,
>
>thank you for your answer. I am using Linux 2.4.19 with the uhci
> and/or the new ehci module (on a single CPU machine). To submit an URB
> continuosly I used the field urb->next this way: urb->next = urb.
I'm not an expert, bu
> Perhaps it would be better to hand parse the structure out?
That's certainly a workable solution. Any structures passed up from
user space to the kernel would likewise need to be similarly hand
parsed. We need to audit the code to see where else this might show up.
Another possibility is to
The device is babbling on the bus... under some situations.
I need to see more logs to be sure what's going on. I need the entire log
from the insertion.
Matt
On Wed, Feb 05, 2003 at 07:06:50PM +1100, Michael Still wrote:
> On Tue, 17 Dec 2002, Matthew Dharm wrote:
>
> > Greg, this patch fixes
Long description follows. the code to review is at the end of the message.
===
I have a device that uses interrupt OUT endpoints and I need a Linux driver.
It seems to me that the existing USB driver stack does not support OUTPUT on
interrupt end-points.
My device is an Ontrak ADU200
On Tue, Feb 04, 2003, Christopher Hoover <[EMAIL PROTECTED]> wrote:
> libusb (cvs development branch) fails on ARM and probably on other RISC
> architectures under Linux because key structures passed from the kernel
> to user space differ in the use of the packed attribute.
>
> In particular struc
On Tuesday 04 February 2003 22:53, David Brownell wrote:
> Duncan Sands wrote:
> > (1) is there any speed advantage to submitting multiple urbs to
> > a bulk endpoint, as compared to using a single urb which is
> > resubmitted in its completion handler?
>
> Yes, most noticeably at high speed where
Let's not forget to update the ATM transmission statistics! And
let's not pretend to support AAL0 when we don't.
speedtouch.c | 11 +++
1 files changed, 3 insertions(+), 8 deletions(-)
diff -Nru a/drivers/usb/misc/speedtouch.c b/drivers/usb/misc/speedtouch.c
--- a/drivers/usb/misc/s
Eliminate a pile of code that isn't used anymore now the new send code
is in place, and make some cosmetic changes.
atmsar.c | 286 ---
atmsar.h |8 -
speedtouch.c | 42 +++-
3 files changed, 15 insertions(+), 321 deletio
Get rid of some unused #defines.
speedtouch.c |3 ---
1 files changed, 3 deletions(-)
diff -Nru a/drivers/usb/misc/speedtouch.c b/drivers/usb/misc/speedtouch.c
--- a/drivers/usb/misc/speedtouch.c Wed Feb 5 16:11:28 2003
+++ b/drivers/usb/misc/speedtouch.c Wed Feb 5 16:11:28 2003
Hi,
this makes sure memory allocation in OHCI for 2.4 always uses either
GFP_ATOMIC or GFP_NOIO. As we cannot tell in 2.4 whether we
are in the block IO path, we always have to assume the worst, hence
GFP_KERNEL cannot be used.
Greg, please apply.
Regards
Oliver
You can i
>> Hello,
>>
>> Is there a way to unlink an URB from within its completition handler
>> without crashing the system? (Both SINC and ASYNC unlink crash the
>> system if doing such a thing)
>Why do you want to unlink an urb from within its completion handler? The
>fact that the handler is r
Am Dienstag, 4. Februar 2003 19:23 schrieb Christopher Li:
> I just did the modify as you said and patch VMware as well.
> And that scanner works fine now.
>
> Will you take this patch?
You need to ask Greg. But even if it were mine to decide, I wouldn't.
If you want to test for 64K, do so. 16*PAG
On Tuesday 04 February 2003 17:04, Duncan Sands wrote:
> (1) is there any speed advantage to submitting multiple urbs to
> a bulk endpoint, as compared to using a single urb which is
> resubmitted in its completion handler?
> (2) is there a noticeable speed difference if you send a buffer
> in many
> I've tried to document the "common" ones in a linux journal article a
> few months ago. But for the others, you'll just have to dig through the
> code, both kernel, and userspace programs.
I found Documentation/ioctl-numbers.txt for new-style ioctls.
Currently I'm using 'N' with a range of 20 t
On Tue, 17 Dec 2002, Matthew Dharm wrote:
> Greg, this patch fixes a stupid coding error on my part. Please apply.
Matthew,
I am using 2.5.59, and am having what seems to me to be a pretty similar
problem to Andre. This patch below is included in 2.5.59 however, and
didn't fix my problem.
To c
34 matches
Mail list logo
