linux-kernel - Re: [PATCH] usb: dwc3: gadget: Add 100uS delay after end transfer command without IOC

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

Message-ID: <20230228032204.mqpaj5toehxkl5tu@synopsys.com>
Date:   Tue, 28 Feb 2023 03:22:19 +0000
From:   Thinh Nguyen <Thinh.Nguyen@...opsys.com>
To:     Wesley Cheng <quic_wcheng@...cinc.com>
CC:     "gregkh@...uxfoundation.org" <gregkh@...uxfoundation.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
        "quic_jackp@...cinc.com" <quic_jackp@...cinc.com>
Subject: Re: [PATCH] usb: dwc3: gadget: Add 100uS delay after end transfer
 command without IOC

On Tue, Feb 28, 2023, Thinh Nguyen wrote:
> On Mon, Feb 27, 2023, Wesley Cheng wrote:
> > Previously, there was a 100uS delay inserted after issuing an end transfer
> > command for specific controller revisions.  This was due to the fact that
> > there was a GUCTL2 bit field which enabled synchronous completion of the
> > end transfer command once the CMDACT bit was cleared in the DEPCMD
> > register.  Since this bit does not exist for all controller revisions, add
> > the delay back in.
> > 
> > An issue was seen where the USB request buffer was unmapped while the DWC3
> > controller was still accessing the TRB.  However, it was confirmed that the
> > end transfer command was successfully submitted. (no end transfer timeout)
> 
> Currently we only check for command active, not completion on teardown.
> 
> > In situations, such as dwc3_gadget_soft_disconnect() and
> > __dwc3_gadget_ep_disable(), the dwc3_remove_request() is utilized, which
> > will issue the end transfer command, and follow up with
> > dwc3_gadget_giveback().  At least for the USB ep disable path, it is
> > required for any pending and started requests to be completed and returned
> > to the function driver in the same context of the disable call.  Without
> > the GUCTL2 bit, it is not ensured that the end transfer is completed before
> > the buffers are unmapped.
> > 
> > Signed-off-by: Wesley Cheng <quic_wcheng@...cinc.com>
> 
> This is expected. We're supposed to make sure the End Transfer command
> complete before accessing the request. Usually on device/endpoint
> teardown, the gadget drivers don't access the stale/incomplete requests
> with -ESHUTDOWN status. There will be problems if we do, and we haven't
> fixed that.
> 
> Adding 100uS may not apply for every device, and we don't need to do
> that for every End Transfer command. Can you try this untested diff
> instead:
> 
> 

Need to make sure ForceRM=0 for the TRB to be updated. Try this instead:

diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
index 3c63fa97a680..55f48fc5844a 100644
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -1686,6 +1686,34 @@ static int __dwc3_gadget_get_frame(struct dwc3 *dwc)
 	return DWC3_DSTS_SOFFN(reg);
 }
 
+static int dwc3_poll_ep_completion(struct dwc3_ep *dep)
+{
+	if (!list_empty(&dep->started_list)) {
+		struct dwc3_request *req;
+		int timeout = 500;
+
+		req = next_request(&dep->started_list);
+		while(--timeout) {
+			/*
+			 * Note: don't check the last enqueued TRB in case
+			 * of short transfer. Check first TRB of a started
+			 * request instead.
+			 */
+			if (!(req->trb->ctrl & DWC3_TRB_CTRL_HWO))
+				break;
+
+			udelay(2);
+		}
+		if (!timeout) {
+			dev_warn(dep->dwc->dev,
+				 "%s is still in-progress\n", dep->name);
+			return -ETIMEDOUT;
+		}
+	}
+
+	return 0;
+}
+
 /**
  * __dwc3_stop_active_transfer - stop the current active transfer
  * @dep: isoc endpoint
@@ -1703,6 +1731,9 @@ static int __dwc3_stop_active_transfer(struct dwc3_ep *dep, bool force, bool int
 	u32 cmd;
 	int ret;
 
+	if (!interrupt)
+		force = false;
+
 	cmd = DWC3_DEPCMD_ENDTRANSFER;
 	cmd |= force ? DWC3_DEPCMD_HIPRI_FORCERM : 0;
 	cmd |= interrupt ? DWC3_DEPCMD_CMDIOC : 0;
@@ -1722,10 +1753,12 @@ static int __dwc3_stop_active_transfer(struct dwc3_ep *dep, bool force, bool int
 	WARN_ON_ONCE(ret);
 	dep->resource_index = 0;
 
-	if (!interrupt)
+	if (!interrupt) {
+		ret = dwc3_poll_ep_completion(dep);
 		dep->flags &= ~DWC3_EP_TRANSFER_STARTED;
-	else if (!ret)
+	} else if (!ret) {
 		dep->flags |= DWC3_EP_END_TRANSFER_PENDING;
+	}
 
 	dep->flags &= ~DWC3_EP_DELAY_STOP;
 	return ret;

Thanks,
Thinh