Hi,
There will be 4 padding bytes in struct fw_cdev_event_response on some platforms
The member:__u32 data will point to these padding bytes. While queue the
response and data in complete_transaction in fw-cdev.c, it will queue like this:
|response(excluding padding bytes)|4 padding bytes|4 padding bytes|data.
It queue 4 extra bytes. That is to say it use "&response + sizeof(response)"
while other place of kernel and userspace library use "&response + offsetof
(typeof(response), data)". So it will lost the last 4 bytes of data.This patch
can fix it while not changing the struct definition.
Sorry for open a new ticket.
Signed-off-by: JiSheng Zhang <[email protected]>
--- old/drivers/firewire/fw-cdev.c
+++ new/drivers/firewire/fw-cdev.c
@@ -382,9 +382,9 @@
response->response.type = FW_CDEV_EVENT_RESPONSE;
response->response.rcode = rcode;
- queue_event(client, &response->event,
- &response->response, sizeof(response->response),
- response->response.data, response->response.length);
+ queue_event(client, &response->event, &response->response,
+ sizeof(response->response) + response->response.length,
+ NULL, 0);
}
static int ioctl_send_request(struct client *client, void *buffer)
JiSheng Zhang wrote at LKML:
> Hi,
>
> There will be 4 padding bytes in struct fw_cdev_event_response on some platforms
> The member:__u32 data will point to these padding bytes. While queue the
> response and data in complete_transaction in fw-cdev.c, it will queue like this:
> |response(excluding padding bytes)|4 padding bytes|4 padding bytes|data.
> It queue 4 extra bytes. That is to say it use "&response + sizeof(response)"
> while other place of kernel and userspace library use "&response + offsetof
> (typeof(response), data)". So it will lost the last 4 bytes of data.This patch
> can fix it while not changing the struct definition.
>
> Sorry for open a new ticket.
>
> Signed-off-by: JiSheng Zhang <[email protected]>
>
> --- old/drivers/firewire/fw-cdev.c
> +++ new/drivers/firewire/fw-cdev.c
> @@ -382,9 +382,9 @@
>
> response->response.type = FW_CDEV_EVENT_RESPONSE;
> response->response.rcode = rcode;
> - queue_event(client, &response->event,
> - &response->response, sizeof(response->response),
> - response->response.data, response->response.length);
> + queue_event(client, &response->event, &response->response,
> + sizeof(response->response) + response->response.length,
> + NULL, 0);
> }
>
> static int ioctl_send_request(struct client *client, void *buffer)
I tested it now on i686, x86-64, and x86-64 with i686 userland, using
firecontrol and gscanbus. As discussed, they got corrupted block read
responses on x86-64 and on x86-64 with i686 userland. The patch fixes this.
I committed it to linux1394-2.6.git#fixes and intend to send it upstream
at the end of the week or so. Thanks for spotting this bug.
One point about which I am not sure about yet is what happens if there
are multiple events queued up before the client can read() them. The
tests which I did so far involved only a single event queued and
dequeued at a time.
PS:
I removed a rule from linux1394-devel's header filters which matched
your previous posts. (Message has priority, but no X-Mailer/User-Agent)
--
Stefan Richter
-=====-==--- -=== =-=--
http://arcgraph.de/sr/
on Sun, 20 Jul 2008 16:00:32 +0200
Stefan Richter <[email protected]> wrote:
> JiSheng Zhang wrote at LKML:
> > Hi,
> >
> > There will be 4 padding bytes in struct fw_cdev_event_response on some platforms
> > The member:__u32 data will point to these padding bytes. While queue the
> > response and data in complete_transaction in fw-cdev.c, it will queue like this:
> > |response(excluding padding bytes)|4 padding bytes|4 padding bytes|data.
> > It queue 4 extra bytes. That is to say it use "&response + sizeof(response)"
> > while other place of kernel and userspace library use "&response + offsetof
> > (typeof(response), data)". So it will lost the last 4 bytes of data.This patch
> > can fix it while not changing the struct definition.
> >
> > Sorry for open a new ticket.
> >
> > Signed-off-by: JiSheng Zhang <[email protected]>
> >
> > --- old/drivers/firewire/fw-cdev.c
> > +++ new/drivers/firewire/fw-cdev.c
> > @@ -382,9 +382,9 @@
> >
> > response->response.type = FW_CDEV_EVENT_RESPONSE;
> > response->response.rcode = rcode;
> > - queue_event(client, &response->event,
> > - &response->response, sizeof(response->response),
> > - response->response.data, response->response.length);
> > + queue_event(client, &response->event, &response->response,
> > + sizeof(response->response) + response->response.length,
> > + NULL, 0);
> > }
> >
> > static int ioctl_send_request(struct client *client, void *buffer)
>
> I tested it now on i686, x86-64, and x86-64 with i686 userland, using
> firecontrol and gscanbus. As discussed, they got corrupted block read
> responses on x86-64 and on x86-64 with i686 userland. The patch fixes this.
>
> I committed it to linux1394-2.6.git#fixes and intend to send it upstream
> at the end of the week or so. Thanks for spotting this bug.
Thanks for committing.
>
> One point about which I am not sure about yet is what happens if there
> are multiple events queued up before the client can read() them. The
> tests which I did so far involved only a single event queued and
> dequeued at a time.
IMHO, even there are multiple events queued, it should work OK because
events are put into event_list rather than a ring buffer, there is no
padding bytes problem between events.
>
>
> PS:
> I removed a rule from linux1394-devel's header filters which matched
> your previous posts. (Message has priority, but no X-Mailer/User-Agent)
Thanks very much
Regards,
JiSheng