linux-kernel - Re: REGRESSION: dwc2: gadget: Add scatter-gather mode

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] [day] [month] [year] [list]

Date:   Thu, 21 Mar 2019 11:36:27 -0700
From:   John Stultz <john.stultz@...aro.org>
To:     Andrzej Pietrasiewicz <andrzejtp2010@...il.com>
Cc:     Marek Szyprowski <m.szyprowski@...sung.com>,
        Felipe Balbi <felipe.balbi@...ux.intel.com>,
        Minas Harutyunyan <Minas.Harutyunyan@...opsys.com>,
        Artur Petrosyan <Arthur.Petrosyan@...opsys.com>,
        Linux USB List <linux-usb@...r.kernel.org>,
        lkml <linux-kernel@...r.kernel.org>,
        Tom Gall <tom.gall@...aro.org>,
        Andrzej Hajda <a.hajda@...sung.com>
Subject: Re: REGRESSION: dwc2: gadget: Add scatter-gather mode

On Thu, Mar 21, 2019 at 12:19 AM Andrzej Pietrasiewicz
<andrzejtp2010@...il.com> wrote:
> Can you please apply this patch:
>
> https://www.spinics.net/lists/linux-usb/msg178195.html
>
> and see if this helps?

Thanks so much for the suggestion!

So I added that, but unfortunately I'm still seeing things crash on boot with:

[   13.655003] init: starting service 'bootanim'...
[   13.699548] functionfs read size 512 > requested size 24, splitting
request into multiple reads.
[   13.699631] ------------[ cut here ]------------
[   13.713062] kernel BUG at mm/slub.c:3950!
[   13.717096] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[   13.722600] Modules linked in:
[   13.725685] CPU: 5 PID: 2123 Comm: adbd Not tainted
5.1.0-rc1-00032-g6cf2064-dirty #1015
[   13.733795] Hardware name: HiKey Development Board (DT)
[   13.739031] pstate: 40400005 (nZcv daif +PAN -UAO)
[   13.743834] pc : kfree+0x210/0x258
[   13.747246] lr : ffs_epfile_io.isra.12+0xfc/0x6a0
[   13.751950] sp : ffffff801199bb50
[   13.755278] x29: ffffff801199bb50 x28: ffffffc074e5c600
[   13.760596] x27: ffffffc06f200e38 x26: 00000000000001e8
[   13.765912] x25: ffffff801199bc98 x24: ffffffc06f0cfa00
[   13.771230] x23: ffffffc07497f000 x22: ffffff801146d000
[   13.776547] x21: ffffff80108a6294 x20: ffffff8012063000
[   13.781865] x19: ffffffbf004818c0 x18: 0000000000000000
[   13.787202] x17: 0000000000000000 x16: ffffffc074e5c600
[   13.792540] x15: 0000000000000000 x14: 0000000000000000
[   13.797877] x13: 0000000000000000 x12: 0000000000000000
[   13.803213] x11: 0000000000000000 x10: 0000000000000000
[   13.808547] x9 : 0000000000000000 x8 : 0000000000000000
[   13.813865] x7 : 0000000000000000 x6 : 000000761e605113
[   13.819182] x5 : 000000761e605113 x4 : 0000000000000000
[   13.824516] x3 : ffffffc06f200e00 x2 : 0000000000000000
[   13.829853] x1 : ffffffbf004818c8 x0 : ffffffbf004818c8
[   13.835199] Process adbd (pid: 2123, stack limit = 0x00000000c418bad6)
[   13.841744] Call trace:
[   13.844208]  kfree+0x210/0x258
[   13.847288]  ffs_epfile_io.isra.12+0xfc/0x6a0
[   13.851673]  ffs_epfile_read_iter+0xb4/0x188
[   13.855973]  new_sync_read+0xe4/0x158
[   13.859643]  __vfs_read+0x2c/0x40
[   13.862977]  vfs_read+0x8c/0x148
[   13.866230]  ksys_read+0x5c/0xc8
[   13.869486]  __arm64_sys_read+0x14/0x20
[   13.873352]  el0_svc_common+0xac/0x108
[   13.877132]  el0_svc_handler+0x2c/0x80
[   13.880910]  el0_svc+0x8/0xc
[   13.883810] Code: f9400260 378000a0 f9400660 37000060 (d4210000)
[   13.889928] ---[ end trace 967312c15ae7c93d ]---


But I like the direction your thinking with that patch.

I realize on some other dwc3 based hardware I was seeing different
issues with larger transfers that cropped up with 772a7a724f69 ("usb:
gadget: f_fs: Allow scatter-gather buffers"), and my hack workaround
there has been to comment out the assignment of
dwc->gadget.sg_supported to true. A similar hack w/ dwc2 seems to
avoid the issue too, so maybe the dwc2 sg_support patch is tripping
over the same thing in the ffs driver?

thanks
-john