linux-kernel - Re: [PATCH] dmaengine: ti: omap-dma: Configure LCH

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]
Date:   Fri, 23 Nov 2018 13:45:46 +0200
From:   Peter Ujfalusi <peter.ujfalusi@...com>
To:     Aaro Koskinen <aaro.koskinen@....fi>
CC:     <vkoul@...nel.org>, <dan.j.williams@...el.com>,
        <dmaengine@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <tony@...mide.com>, <linux-omap@...r.kernel.org>,
        <rmk+kernel@...linux.org.uk>,
        Felipe Balbi <felipe.balbi@...ux.intel.com>
Subject: Re: [PATCH] dmaengine: ti: omap-dma: Configure LCH_TYPE for OMAP1



On 23/11/2018 0.01, Aaro Koskinen wrote:
> Hi,
> 
> On Thu, Nov 22, 2018 at 10:31:31AM +0200, Peter Ujfalusi wrote:
>> On 20/11/2018 23.04, Aaro Koskinen wrote:
>>> On Tue, Nov 20, 2018 at 09:28:37AM +0200, Peter Ujfalusi wrote:
>>>> On 19/11/2018 20.46, Aaro Koskinen wrote:
>>>>> On Mon, Nov 19, 2018 at 12:40:40PM +0200, Peter Ujfalusi wrote:
>>>>>> When the channel is configured for slave operation the LCH_TYPE needs to be
>>>>>> set to LCh-P. For memcpy channels the LCH_TYPE must be set to LCh-2D.
>>>>>>
>>>>>> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@...com>
>>>>>
>>>>> I don't have the documentation, but based on what omap_udc driver (still
>>>>> using the legacy OMAP DMA API) does this seems to be correct.
>>>>
>>>> They are hard to fine, true. From the omap1710 TRM:
>>>>
>>>> Logical channel types (LCh types) supported are:
>>>> - LCh-2D for nonsynchronized transfers (memory transfers, 1D and 2D)
>>>> - LCh-P for synchronized transfers (mostly peripheral transfers)
>>>> - LCh-PD similar to LCh-P but runs on a dedicated physical channel
>>>> - LCh-G for graphical transfers/operations
>>>> - LCh-D for display transfers
>>>
>>> (I found a public document "OMAP5912 Multimedia Processor Direct
>>> Memory Access (DMA) Support Reference Guide", documenting these; easy
>>> to confuse with "OMAP5910 Dual-Core Processor System DMA Controller
>>> Reference Guide".)
>>>
>>>> Looking at other part it looks like hat LCH-2D channel mode can happily
>>>> service a peripheral. LCH-P supports the same features as LCH-2D, but it
>>>> lacks support for Single/Double-indexed addressing mode on the
>>>> peripheral port side.
>>>>
>>>> So, this patch might not be needed at all. Can you test the omap_udc
>>>> with s/OMAP_DMA_LCH_P/OMAP_DMA_LCH_2D
>>>>
>>>> If USB works, then we can just drop this patch.
>>>
>>> Unfortunately omap_udc does not seem to work at all anymore with DMA on
>>> my 770 setup. :-(
>>>
>>> I had switched to PIO mode in 2015 since the WARNs about legacy DMA
>>> API were too annoying and flooding the console. And now that I tried
>>> using DMA again with g_ether, it doesn't work anymore. The device get's
>>> recognized on host side, but no traffic goes through. Switching back to
>>> PIO makes it to work again.
>>
>> After some tinkering I got omap_udc working with DMA (not DMAengine,
>> yet) on 770 - using nfsroot. Configuring the channels to OMAP_DMA_LCH_2D
>> works as expected.
> 
> Would be interesting to know how you got it working with DMA. Which
> gadget driver were you using?
> 
> I bisected my issue, and got:
> 
> commit 387f869d2579e379ee343f5493dcd360be60f5c6 (refs/bisect/bad)
> Author: Felipe Balbi <felipe.balbi@...ux.intel.com>
> Date:   Wed Mar 22 13:25:18 2017 +0200
> 
>     usb: gadget: u_ether: conditionally align transfer size

I just:
commit 0d61d79625202c1c4fcf07fb960e27984a3657a3
Author: Peter Ujfalusi <peter.ujfalusi@...com>
Date:   Thu Nov 22 10:36:55 2018 +0200

    usb: gadget: omap_udc: HACK: Make RX dma work
    
    partial packets do work...
    
    Signed-off-by: Peter Ujfalusi <peter.ujfalusi@...com>

diff --git a/drivers/usb/gadget/udc/omap_udc.c b/drivers/usb/gadget/udc/omap_udc.c
index 94128eb69d97..0748c3841a25 100644
--- a/drivers/usb/gadget/udc/omap_udc.c
+++ b/drivers/usb/gadget/udc/omap_udc.c
@@ -886,14 +886,14 @@ omap_ep_queue(struct usb_ep *_ep, struct usb_request *_req, gfp_t gfp_flags)
 
        /* this isn't bogus, but OMAP DMA isn't the only hardware to
         * have a hard time with partial packet reads...  reject it.
+        * Wait a minute, it does work :o
         */
        if (use_dma
                        && ep->has_dma
                        && ep->bEndpointAddress != 0
                        && (ep->bEndpointAddress & USB_DIR_IN) == 0
                        && (req->req.length % ep->ep.maxpacket) != 0) {
-               DBG("%s, no partial packet OUT reads\n", __func__);
-               return -EMSGSIZE;
+               DBG("%s, partial packet OUT, might not work?\n", __func__);
        }
 
        udc = ep->udc;

> With that reverted, the DMA works OK (and I can also now confirm that
> OMAP_DMA_LCH_2D works). I haven't yet checked if we actually need that
> quirk in OMAP UDC,

The omap_udc driver is a bit of a mess, need to check it myself, but for
now we can just set the quirk_ep_out_aligned_size and investigate later.

> or if this is related to RMK's findings of potential
> bugs in the driver. Anyway, there is clearly yet another regression.

I'll check Russell's mail.

> 
> A.
> 

- Péter

Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki. Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki