linux-kernel - RE: [RFC][PATCH 2/2] usb: dwc3: gadget: Correct the logic for finding last SG entry

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: <BYAPR02MB55910511FDAA7FDA0C647ABEA70F0@BYAPR02MB5591.namprd02.prod.outlook.com>
Date:   Thu, 23 Jan 2020 15:50:17 +0000
From:   Anurag Kumar Vulisha <anuragku@...inx.com>
To:     Felipe Balbi <balbi@...nel.org>,
        John Stultz <john.stultz@...aro.org>,
        lkml <linux-kernel@...r.kernel.org>
CC:     Felipe Balbi <felipe.balbi@...ux.intel.com>,
        Yang Fei <fei.yang@...el.com>,
        Thinh Nguyen <thinhn@...opsys.com>,
        Tejas Joglekar <tejas.joglekar@...opsys.com>,
        Andrzej Pietrasiewicz <andrzej.p@...labora.com>,
        Jack Pham <jackp@...eaurora.org>, Todd Kjos <tkjos@...gle.com>,
        Greg KH <gregkh@...uxfoundation.org>,
        Linux USB List <linux-usb@...r.kernel.org>,
        stable <stable@...r.kernel.org>,
        John Stultz <john.stultz@...aro.org>
Subject: RE: [RFC][PATCH 2/2] usb: dwc3: gadget: Correct the logic for finding
 last SG entry

Hi Felipe,
>-----Original Message-----
>From: Felipe Balbi <balbif@...il.com> On Behalf Of Felipe Balbi
>Sent: Thursday, January 23, 2020 12:56 PM
>To: John Stultz <john.stultz@...aro.org>; lkml <linux-kernel@...r.kernel.org>
>Cc: Anurag Kumar Vulisha <anuragku@...inx.com>; Felipe Balbi
><felipe.balbi@...ux.intel.com>; Yang Fei <fei.yang@...el.com>; Thinh Nguyen
><thinhn@...opsys.com>; Tejas Joglekar <tejas.joglekar@...opsys.com>;
>Andrzej Pietrasiewicz <andrzej.p@...labora.com>; Jack Pham
><jackp@...eaurora.org>; Todd Kjos <tkjos@...gle.com>; Greg KH
><gregkh@...uxfoundation.org>; Linux USB List <linux-usb@...r.kernel.org>;
>stable <stable@...r.kernel.org>; John Stultz <john.stultz@...aro.org>
>Subject: Re: [RFC][PATCH 2/2] usb: dwc3: gadget: Correct the logic for finding
>last SG entry
>
>
>Hi,
>
>John Stultz <john.stultz@...aro.org> writes:
>> From: Anurag Kumar Vulisha <anurag.kumar.vulisha@...inx.com>
>>
>> As a process of preparing TRBs usb_gadget_map_request_by_dev() is
>> called from dwc3_prepare_trbs() for mapping the request. This will
>> call dma_map_sg() if req->num_sgs are greater than 0. dma_map_sg()
>> will map the sg entries in sglist and return the number of mapped SGs.
>> As a part of mapping, some sg entries having contigous memory may be
>> merged together into a single sg (when IOMMU used). So, the number of
>> mapped sg entries may not be equal to the number of orginal sg entries
>> in the request (req->num_sgs).
>>
>> As a part of preparing the TRBs, dwc3_prepare_one_trb_sg() iterates
>> over the sg entries present in the sglist and calls sg_is_last() to
>> identify whether the sg entry is last and set IOC bit for the last sg
>> entry. The
>> sg_is_last() determines last sg if SG_END is set in sg->page_link.
>> When IOMMU used, dma_map_sg() merges 2 or more sgs into a single sg
>> and it doesn't retain the page_link properties. Because of this reason
>> the
>> sg_is_last() may not find SG_END and thus resulting in IOC bit never
>> getting set.
>>
>> For example:
>>
>> Consider a request having 8 sg entries with each entry having a length
>> of
>> 4096 bytes. Assume that sg1 & sg2, sg3 & sg4, sg5 & sg6, sg7 & sg8 are
>> having contigous memory regions.
>>
>> Before calling dma_map_sg():
>>             sg1-->sg2-->sg3-->sg4-->sg6-->sg7-->sg8
>> dma_length: 4K    4K    4K    4K    4K    4K    4K
>> SG_END:     False False False False False False True
>> num_sgs = 8
>> num_mapped_sgs = 0
>>
>> The dma_map_sg() merges sg1 & sg2 memory regions into sg1->dma_address.
>> Similarly sg3 & sg4 into sg2->dma_address, sg5 & sg6 into the
>> sg3->dma_address and sg6 & sg8 into sg4->dma_address. Here the memory
>> regions are merged but the page_link properties like SG_END are not
>> retained into the merged sgs.
>>
>> After calling dma_map_sg();
>>             sg1-->sg2-->sg3-->sg4-->sg6-->sg7-->sg8
>> dma_length: 8K    8K    8K    8K    0K    0K     0K
>> SG_END:     False False False False False False True
>> num_sgs = 8
>> num_mapped_sgs = 4
>>
>> After calling dma_map_sg(), sg1,sg2,sg3,sg4 are having dma_length of
>> 8096 bytes each and remaining sg4,sg5,sg6,sg7 are having 0 bytes of
>> dma_length.
>>
>> After dma_map_sg() is performed dma_perpare_trb_sg() iterates on all
>> sg entries and sets IOC bit only for the sg8 (since sg_is_last()
>> returns true only for sg8). But after calling dma_map_sg() the valid
>> data are present only till sg4 and the IOC bit should be set for sg4
>> TRB only (which is not happening in the present code)
>>
>> The above mentioned issue can be fixed by determining last sg based on
>> the
>> req->num_queued_sgs instead of sg_is_last(). If (req->num_queued_sgs +
>> req->1)
>> is equal to req->num_mapped_sgs, then this sg is the last sg. In the
>> above example, the dwc3 driver has already queued 3 sgs (upto sg3), so
>> the num_queued_sgs = 3. On preparing the next sg (i.e sg4), check for
>> last sg (num_queued_sgs + 1) == num_mapped_sgs becomes true. So, the
>> driver sets IOC bit for sg4. This patch does the same.
>>
>> At a practical level, this patch resolves USB transfer stalls seen
>> with adb on dwc3 based db845c, pixel3 and other qcom hardware after
>> functionfs gadget added scatter-gather support around v4.20.
>>
>> Cc: Felipe Balbi <felipe.balbi@...ux.intel.com>
>> Cc: Yang Fei <fei.yang@...el.com>
>> Cc: Thinh Nguyen <thinhn@...opsys.com>
>> Cc: Tejas Joglekar <tejas.joglekar@...opsys.com>
>> Cc: Andrzej Pietrasiewicz <andrzej.p@...labora.com>
>> Cc: Jack Pham <jackp@...eaurora.org>
>> Cc: Todd Kjos <tkjos@...gle.com>
>> Cc: Greg KH <gregkh@...uxfoundation.org>
>> Cc: Linux USB List <linux-usb@...r.kernel.org>
>> Cc: stable <stable@...r.kernel.org>
>> Signed-off-by: Anurag Kumar Vulisha <anurag.kumar.vulisha@...inx.com>
>> [jstultz: Add note to end of commit message on specific issue this
>> resovles]
>> Signed-off-by: John Stultz <john.stultz@...aro.org>
>> ---
>>  drivers/usb/dwc3/gadget.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>> index 1edce3bbb55c..30a80bc97cfe 100644
>> --- a/drivers/usb/dwc3/gadget.c
>> +++ b/drivers/usb/dwc3/gadget.c
>> @@ -1068,7 +1068,7 @@ static void dwc3_prepare_one_trb_sg(struct
>dwc3_ep *dep,
>>  		unsigned int rem = length % maxp;
>>  		unsigned chain = true;
>>
>> -		if (sg_is_last(s))
>> +		if ((req->num_queued_sgs + 1) == req-
>>request.num_mapped_sgs)
>
>This is probably a bug on DMA API. If it combines pages from scatter-list, then it
>should also move the last SG so sg_is_last() continues to work.
>
>I had asked author to discuss this with DMA API maintainers. Can you do that?
>
I was stuck with other tasks, so couldn't discuss with DMA maintainers on this.
I am sorry for that. 

Hi John,
Thanks for bringing this patch back . Please let me know if I can help you with
anything on this. If you want, I am ready to start working on this.

Thanks,
Anurag Kumar Vulisha