Commit 6c49dc8c authored by David Brownell's avatar David Brownell Committed by Greg Kroah-Hartman

[PATCH] usbtest, covers control queueing and fault cleanup

I wrote this a while back, finally debugged it.  This covers
some functionality that 2.5 newly demands of all HCDs:  control
requests can be queued.  (Example:  a user mode driver can talk
on one interface, and a kernel mode one can talk on another,
no need to handshake about who can make control requests.)

The good news is that all the HCDs seem (light testing) to do
the right things ... until some of the requests (intentionally)
trigger routine faults (like protocol stalls) which the HCDs
need to recover from.  At that point, uhci-hcd started acting
confused (it's got newish queueing code); details will come
separately.  The other two HCDs acted fine.  I had expected more
trouble there, maybe it'll show up later on.
parent 6da387fb
...@@ -546,7 +546,300 @@ static int ch9_postconfig (struct usbtest_dev *dev) ...@@ -546,7 +546,300 @@ static int ch9_postconfig (struct usbtest_dev *dev)
/*-------------------------------------------------------------------------*/ /*-------------------------------------------------------------------------*/
// control queueing !! /* use ch9 requests to test whether:
* (a) queues work for control, keeping N subtests queued and
* active (auto-resubmit) for M loops through the queue.
* (b) protocol stalls (control-only) will autorecover.
* it's quite not like bulk/intr; no halt clearing.
* (c) short control reads are reported and handled.
*/
struct ctrl_ctx {
spinlock_t lock;
struct usbtest_dev *dev;
struct completion complete;
unsigned count;
unsigned pending;
int status;
struct urb **urb;
struct usbtest_param *param;
};
static void ctrl_complete (struct urb *urb, struct pt_regs *regs)
{
struct ctrl_ctx *ctx = urb->context;
struct usb_ctrlrequest *reqp;
int status = urb->status;
reqp = (struct usb_ctrlrequest *)urb->setup_packet;
spin_lock (&ctx->lock);
ctx->count--;
ctx->pending--;
/* FIXME verify that the completions are in the right sequence.
* we could store the test number with the setup packet, that
* buffer has extra space.
*/
switch (status) {
case 0: /* success */
case -EREMOTEIO: /* short read */
if (reqp->bRequestType == (USB_DIR_IN|USB_RECIP_DEVICE)
&& reqp->bRequest == USB_REQ_GET_DESCRIPTOR
&& ((le16_to_cpu (reqp->wValue) >> 8)
== USB_DT_DEVICE)) {
if (reqp->wLength > USB_DT_DEVICE_SIZE
&& status == -EREMOTEIO)
status = 0;
else if (reqp->wLength == USB_DT_DEVICE_SIZE
&& status != 0)
status = -EIO;
if (status)
goto error;
}
break;
case -ECONNRESET: /* async unlink */
break;
case -EPIPE: /* (protocol) stall */
if (reqp->bRequestType == (USB_DIR_IN|USB_RECIP_INTERFACE)
&& reqp->bRequest == USB_REQ_GET_INTERFACE)
status = 0;
else if (reqp->bRequestType == (USB_DIR_IN|USB_RECIP_DEVICE)
&& reqp->bRequest == USB_REQ_GET_DESCRIPTOR) {
switch (le16_to_cpu (reqp->wValue) >> 8) {
case USB_DT_DEVICE_QUALIFIER:
case USB_DT_OTHER_SPEED_CONFIG:
case USB_DT_INTERFACE:
case USB_DT_ENDPOINT:
status = 0;
}
} else if (reqp->bRequestType == USB_RECIP_ENDPOINT
&& reqp->bRequest == USB_REQ_CLEAR_FEATURE)
status = 0;
/* some stalls we plan on; others would be errors */
if (status == 0)
break;
/* else FALLTHROUGH */
error:
default: /* this fault's an error */
if (ctx->status == 0) {
int i;
ctx->status = status;
info ("control queue %02x.%02x, err %d, %d left",
reqp->bRequestType, reqp->bRequest,
status, ctx->count);
/* FIXME use this "unlink everything" exit route
* in all cases, not just for fault cleanup.
* it'll be another test mode, but one that makes
* testing be more consistent.
*/
/* unlink whatever's still pending */
for (i = 0; i < ctx->param->sglen; i++) {
struct urb *u = ctx->urb [i];
if (u == urb || !u->dev)
continue;
status = usb_unlink_urb (u);
switch (status) {
case -EINPROGRESS:
case -EBUSY:
continue;
default:
dbg ("urb unlink --> %d", status);
}
}
status = ctx->status;
}
}
/* resubmit if we need to, else mark this as done */
if ((status == 0) && (ctx->pending < ctx->count)) {
if ((status = usb_submit_urb (urb, SLAB_ATOMIC)) != 0) {
dbg ("can't resubmit ctrl %02x.%02x, err %d",
reqp->bRequestType, reqp->bRequest, status);
urb->dev = 0;
} else
ctx->pending++;
} else
urb->dev = 0;
/* signal completion when nothing's queued */
if (ctx->pending == 0)
complete (&ctx->complete);
spin_unlock (&ctx->lock);
}
static int
test_ctrl_queue (struct usbtest_dev *dev, struct usbtest_param *param)
{
struct usb_device *udev = testdev_to_usbdev (dev);
struct urb **urb;
struct ctrl_ctx context;
int i;
spin_lock_init (&context.lock);
context.dev = dev;
init_completion (&context.complete);
context.count = param->sglen * param->iterations;
context.pending = 0;
context.status = -ENOMEM;
context.param = param;
/* allocate and init the urbs we'll queue.
* as with bulk/intr sglists, sglen is the queue depth; it also
* controls which subtests run (more tests than sglen) or rerun.
*/
urb = kmalloc (param->sglen * sizeof (struct urb *), SLAB_KERNEL);
if (!urb)
goto cleanup;
memset (urb, 0, param->sglen * sizeof (struct urb *));
for (i = 0; i < param->sglen; i++) {
int pipe = usb_rcvctrlpipe (udev, 0);
unsigned len;
struct urb *u;
struct usb_ctrlrequest req, *reqp;
/* requests here are mostly expected to succeed on any
* device, but some are chosen to trigger protocol stalls
* or short reads.
*/
memset (&req, 0, sizeof req);
req.bRequest = USB_REQ_GET_DESCRIPTOR;
req.bRequestType = USB_DIR_IN|USB_RECIP_DEVICE;
switch (i % 12 /* number of subtest cases here */) {
case 0: // get device descriptor
req.wValue = cpu_to_le16 (USB_DT_DEVICE << 8);
len = sizeof (struct usb_device_descriptor);
break;
case 1: // get first config descriptor (only)
req.wValue = cpu_to_le16 ((USB_DT_CONFIG << 8) | 0);
len = sizeof (struct usb_config_descriptor);
break;
case 2: // get altsetting (OFTEN STALLS)
req.bRequest = USB_REQ_GET_INTERFACE;
req.bRequestType = USB_DIR_IN|USB_RECIP_INTERFACE;
// index = 0 means first interface
len = 1;
break;
case 3: // get interface status
req.bRequest = USB_REQ_GET_STATUS;
req.bRequestType = USB_DIR_IN|USB_RECIP_INTERFACE;
// interface 0
len = 2;
break;
case 4: // get device status
req.bRequest = USB_REQ_GET_STATUS;
req.bRequestType = USB_DIR_IN|USB_RECIP_DEVICE;
len = 2;
break;
case 5: // get device qualifier (MAY STALL)
req.wValue = cpu_to_le16 (USB_DT_DEVICE_QUALIFIER << 8);
len = sizeof (struct usb_qualifier_descriptor);
break;
case 6: // get first config descriptor, plus interface
req.wValue = cpu_to_le16 ((USB_DT_CONFIG << 8) | 0);
len = sizeof (struct usb_config_descriptor);
len += sizeof (struct usb_interface_descriptor);
break;
case 7: // get interface descriptor (ALWAYS STALLS)
req.wValue = cpu_to_le16 (USB_DT_INTERFACE << 8);
// interface == 0
len = sizeof (struct usb_interface_descriptor);
break;
// NOTE: two consecutive stalls in the queue here.
// that tests fault recovery a bit more aggressively.
case 8: // clear endpoint halt (USUALLY STALLS)
req.bRequest = USB_REQ_CLEAR_FEATURE;
req.bRequestType = USB_RECIP_ENDPOINT;
// wValue 0 == ep halt
// wIndex 0 == ep0 (shouldn't halt!)
len = 0;
pipe = usb_sndctrlpipe (udev, 0);
break;
case 9: // get endpoint status
req.bRequest = USB_REQ_GET_STATUS;
req.bRequestType = USB_DIR_IN|USB_RECIP_ENDPOINT;
// endpoint 0
len = 2;
break;
case 10: // trigger short read (EREMOTEIO)
req.wValue = cpu_to_le16 ((USB_DT_CONFIG << 8) | 0);
len = 1024;
break;
// NOTE: two consecutive _different_ faults in the queue.
case 11: // get endpoint descriptor (ALWAYS STALLS)
req.wValue = cpu_to_le16 (USB_DT_ENDPOINT << 8);
// endpoint == 0
len = sizeof (struct usb_interface_descriptor);
break;
// NOTE: sometimes even a third fault in the queue!
case 12: // get string 0 descriptor (MAY STALL)
req.wValue = cpu_to_le16 (USB_DT_STRING << 8);
// string == 0, for language IDs
len = sizeof (struct usb_interface_descriptor);
break;
default:
err ("bogus number of ctrl queue testcases!");
context.status = -EINVAL;
goto cleanup;
}
req.wLength = cpu_to_le16 (len);
urb [i] = u = simple_alloc_urb (udev, pipe, len);
if (!u)
goto cleanup;
reqp = usb_buffer_alloc (udev, sizeof req, SLAB_KERNEL,
&u->setup_dma);
if (!reqp)
goto cleanup;
*reqp = req;
u->setup_packet = (char *) reqp;
u->context = &context;
u->complete = ctrl_complete;
u->transfer_flags |= URB_ASYNC_UNLINK;
}
/* queue the urbs */
context.urb = urb;
spin_lock_irq (&context.lock);
for (i = 0; i < param->sglen; i++) {
context.status = usb_submit_urb (urb [i], SLAB_ATOMIC);
if (context.status != 0) {
dbg ("can't submit urb[%d], status %d",
i, context.status);
context.count = context.pending;
break;
}
context.pending++;
}
spin_unlock_irq (&context.lock);
/* FIXME set timer and time out; provide a disconnect hook */
/* wait for the last one to complete */
wait_for_completion (&context.complete);
cleanup:
for (i = 0; i < param->sglen; i++) {
if (!urb [i])
continue;
urb [i]->dev = udev;
if (urb [i]->setup_packet)
usb_buffer_free (udev, sizeof (struct usb_ctrlrequest),
urb [i]->setup_packet,
urb [i]->setup_dma);
simple_free_urb (urb [i]);
}
kfree (urb);
return context.status;
}
/*-------------------------------------------------------------------------*/ /*-------------------------------------------------------------------------*/
...@@ -830,7 +1123,16 @@ usbtest_ioctl (struct usb_interface *intf, unsigned int code, void *buf) ...@@ -830,7 +1123,16 @@ usbtest_ioctl (struct usb_interface *intf, unsigned int code, void *buf)
dbg ("ch9 subset failed, iterations left %d", i); dbg ("ch9 subset failed, iterations left %d", i);
break; break;
// case 10: queued control /* queued control messaging */
case 10:
if (param->sglen == 0)
break;
retval = 0;
dbg ("%s TEST 10: queue %d control calls, %d times",
dev->id, param->sglen,
param->iterations);
retval = test_ctrl_queue (dev, param);
break;
/* simple non-queued unlinks (ring with one urb) */ /* simple non-queued unlinks (ring with one urb) */
case 11: case 11:
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment