linux-kernel - [PATCH] b: re-queue tx dma request on herror

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <1479282123-14316-1-git-send-email-muvarov@gmail.com>
Date:   Wed, 16 Nov 2016 10:42:03 +0300
From:   Max Uvarov <muvarov@...il.com>
To:     linux-kernel@...r.kernel.org
Cc:     linux-usb@...r.kernel.org, gregkh@...uxfoundation.org,
        b-liu@...com, Max Uvarov <muvarov@...il.com>
Subject: [PATCH] b: re-queue tx dma request on herror

Some times dma transfer to usb endpoint fails:

[ 78.378283] musb-hdrc musb-hdrc.1.auto: Start TX10 dma
[ 78.410763] musb-hdrc musb-hdrc.1.auto: OUT/TX10 end, csr 3400, dma
[ 78.410896] musb-hdrc musb-hdrc.1.auto: complete dc01eb00 usb_api_blocking_completion+0x0/0x24 [usbcore] (0), dev4 ep1out, 10/10
[ 78.411181] musb-hdrc musb-hdrc.1.auto: qh dc01ed00 periodic slot 10
[ 78.411205] musb-hdrc musb-hdrc.1.auto: qh dc01ed00 urb dc01eb00 dev4 ep1out-intr, hw_ep 10, dd624d00/10
[ 78.411223] musb-hdrc musb-hdrc.1.auto: --> hw10 urb dc01eb00 spd2 dev4 ep1out h_addr83 h_port02 bytes 10
[ 78.411244] musb-hdrc musb-hdrc.1.auto: check whether there's still time for periodic Tx
[ 78.411256] musb-hdrc musb-hdrc.1.auto: Start TX10 dma
[ 78.443762] musb-hdrc musb-hdrc.1.auto: OUT/TX10 end, csr 3500, dma

success transmition:

[ 78.443889] musb-hdrc musb-hdrc.1.auto: complete dc01eb00 usb_api_blocking_completion+0x0/0x24 [usbcore] (0), dev4 ep1out, 10/10
[ 78.444170] musb-hdrc musb-hdrc.1.auto: qh dc01ed00 periodic slot 10
[ 78.444195] musb-hdrc musb-hdrc.1.auto: qh dc01ed00 urb dc01eb00 dev4 ep1out-intr, hw_ep 10, dd624d00/10
[ 78.444213] musb-hdrc musb-hdrc.1.auto: --> hw10 urb dc01eb00 spd2 dev4 ep1out h_addr83 h_port02 bytes 10
[ 78.444232] musb-hdrc musb-hdrc.1.auto: check whether there's still time for periodic Tx
[ 78.444245] musb-hdrc musb-hdrc.1.auto: Start TX10 dma
[ 78.540761] musb-hdrc musb-hdrc.1.auto: OUT/TX10 end, csr 3504, dma

failed transmission:

[ 78.540780] musb-hdrc musb-hdrc.1.auto: TX 3strikes on ep=10 set ETIMEDOUT
[ 78.540897] musb-hdrc musb-hdrc.1.auto: complete dc01eb00 usb_api_blocking_completion+0x0/0x24 [usbcore] (-110), dev4 ep1out, 10/10
[ 78.540945] musb-hdrc musb-hdrc.1.auto: extra TX10 ready, csr 2500
[ 78.540989] usb 2-1.1.2: urb wait ok but retval -110

Use to reproduce is writes to /dev/hidraw0 device which end up with early
unexpected timeout and -110 errno set.

Code sets timeout to 5 seconds
vfs_write()->hidraw_write()->hidraw_send_report()->usbhid_output_report()
   ret = usb_interrupt_msg(dev, usbhid->urbout->pipe,
                           buf, count, &actual_length,
                           USB_CTRL_SET_TIMEOUT);
where is set to 5 second:

when it wents to usb_start_wait_urb() where waits for completition:
   retval = usb_submit_urb(urb, GFP_NOIO);

   expire = timeout ? msecs_to_jiffies(timeout) : MAX_SCHEDULE_TIMEOUT;
   if (!wait_for_completion_timeout(&ctx.done, expire)) {

So user space application expects that write will be done in 5 seconds or
error will happen. But musb driver exists this logic on first dma error
without trying to retransmit current urb. This patch adds current request
to the end of list, destroys current dma transfer and renew transmission.
In that case this urb transmitted in next cycle and not failing with error
before timeout.

Signed-off-by: Max Uvarov <muvarov@...il.com>
---
 drivers/usb/musb/musb_host.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/drivers/usb/musb/musb_host.c b/drivers/usb/musb/musb_host.c
index 53bc4ce..e44ae95 100644
--- a/drivers/usb/musb/musb_host.c
+++ b/drivers/usb/musb/musb_host.c
@@ -1293,11 +1293,11 @@ void musb_host_tx(struct musb *musb, u8 epnum)
 		status = -EPIPE;
 
 	} else if (tx_csr & MUSB_TXCSR_H_ERROR) {
-		/* (NON-ISO) dma was disabled, fifo flushed */
 		musb_dbg(musb, "TX 3strikes on ep=%d", epnum);
-
-		status = -ETIMEDOUT;
-
+		if (dma)
+			musb_bulk_nak_timeout(musb, hw_ep, 0);
+		else
+			status = -ETIMEDOUT;
 	} else if (tx_csr & MUSB_TXCSR_H_NAKTIMEOUT) {
 		if (USB_ENDPOINT_XFER_BULK == qh->type && qh->mux == 1
 				&& !list_is_singular(&musb->out_bulk)) {
-- 
1.9.1