diff mbox

usb: dwc2: fix panic for coherent memory allocation

Message ID 1483671005-13394-1-git-send-email-leo.yan@linaro.org
State New
Headers show

Commit Message

Leo Yan Jan. 6, 2017, 2:50 a.m. UTC
When use configfs to configure USB port as as ethernet gadget, the
kernel has panic with below backtrace; it clearly indicates the
coherent memory allocation is happened in the interrupt context,
but the function __get_vm_area_node() is possible to be scheduling
out so function __get_vm_area_node() triggers panic if it's called
from interrupt context.

To fix this issue, simply to change gfp_flag from GFP_KERNEL to
GFP_NOWAIT so it can support to allocate coherent memory from
interrupt context.

[    8.036567] ------------[ cut here ]------------
[    8.041188] kernel BUG at mm/vmalloc.c:1344!
[    8.045462] Internal error: Oops - BUG: 0 [#1] PREEMPT SMP
[    8.050950] Modules linked in:
[    8.054007] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G        W       4.10.0-rc2-00201-ged40875 #645
[    8.062978] Hardware name: HiKey Development Board (DT)
[    8.068206] task: ffff000008dc2900 task.stack: ffff000008db0000
[    8.176374] Process swapper/0 (pid: 0, stack limit = 0xffff000008db0000)
[    8.183080] Stack: (0xffff80003df19a80 to 0xffff000008db4000)
[    8.188829] Call trace:
[    8.191276] Exception stack(0xffff80003df198b0 to 0xffff80003df199e0)
[    8.197722] 98a0:                                   000000000002d048 0001000000000000
[    8.205564] 98c0: ffff80003df19a80 ffff0000081aca98 00000000000001c0 0000000000000004
[    8.213407] 98e0: ffff7e0000b48000 0000000000000009 ffff000008e93000 ffff7e0000b41200
[    8.221250] 9900: ffff80003df19950 ffff000008174e40 0000000000000000 ffff000008e93000
[    8.229093] 9920: ffff000008e92b00 ffff000008e92000 ffff000008c37000 ffff000008e92b00
[    8.236935] 9940: 00000000000001c0 0000000000000004 0000000000001000 0000000000010002
[    8.244777] 9960: 0000000000000008 ffff000008000000 ffff7dffbfff0000 00000000024000c0
[    8.252619] 9980: 0000000000000000 0000000000000001 ffff7e0000000000 000000000002d048
[    8.260462] 99a0: fffffffffffffe00 0000000000000001 0000000000000068 dead000000000200
[    8.268305] 99c0: dead000000000100 00000000000001c4 fffffffffffffffe 0000000000000000
[    8.276148] [<ffff0000081aca98>] __get_vm_area_node.isra.37+0x180/0x184
[    8.282769] [<ffff0000081acbc8>] get_vm_area_caller+0x34/0x3c
[    8.288522] [<ffff000008536c4c>] dma_common_pages_remap+0x28/0x70
[    8.294621] [<ffff000008536d60>] dma_common_contiguous_remap+0xcc/0xf0
[    8.301154] [<ffff000008097308>] __dma_alloc+0x10c/0x228
[    8.306471] [<ffff00000869f6e0>] dwc2_hsotg_ep_enable+0x360/0x4c8
[    8.312571] [<ffff0000086e17a8>] usb_ep_enable+0x3c/0x5c
[    8.317888] [<ffff0000086e33e4>] ecm_set_alt+0xac/0x164
[    8.323117] [<ffff0000086dcbb8>] composite_setup+0xbac/0x140c
[    8.328868] [<ffff0000086a02c4>] dwc2_hsotg_complete_setup+0xb8/0x368
[    8.335315] [<ffff0000086e0600>] usb_gadget_giveback_request+0x10/0x18
[    8.341849] [<ffff00000869f00c>] dwc2_hsotg_complete_request+0x7c/0x130
[    8.348470] [<ffff00000869ffe4>] dwc2_hsotg_handle_outdone+0xfc/0x1e0
[    8.354917] [<ffff0000086a1500>] dwc2_hsotg_epint+0x638/0x844
[    8.360667] [<ffff0000086a1e54>] dwc2_hsotg_irq+0x374/0x6f4
[    8.366246] [<ffff000008109424>] __handle_irq_event_percpu+0xa0/0x128
[    8.372693] [<ffff0000081094c8>] handle_irq_event_percpu+0x1c/0x54
[    8.378879] [<ffff000008109544>] handle_irq_event+0x44/0x74
[    8.384456] [<ffff00000810cbf4>] handle_fasteoi_irq+0xb4/0x184
[    8.390294] [<ffff000008108548>] generic_handle_irq+0x24/0x38
[    8.396045] [<ffff000008108bb8>] __handle_domain_irq+0x60/0xac
[    8.401883] [<ffff00000808161c>] gic_handle_irq+0x54/0xa0

Signed-off-by: Leo Yan <leo.yan@linaro.org>

---
 drivers/usb/dwc2/gadget.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

-- 
2.7.4

Comments

John Stultz Jan. 6, 2017, 9:36 p.m. UTC | #1
On Thu, Jan 5, 2017 at 6:50 PM, Leo Yan <leo.yan@linaro.org> wrote:
> When use configfs to configure USB port as as ethernet gadget, the

> kernel has panic with below backtrace; it clearly indicates the

> coherent memory allocation is happened in the interrupt context,

> but the function __get_vm_area_node() is possible to be scheduling

> out so function __get_vm_area_node() triggers panic if it's called

> from interrupt context.

>

> To fix this issue, simply to change gfp_flag from GFP_KERNEL to

> GFP_NOWAIT so it can support to allocate coherent memory from

> interrupt context.


This looks like it duplicates the similar patch I'm told should be
queued for 4.10-rc3:
    https://lkml.org/lkml/2016/12/1/120

thanks
-john
Leo Yan Jan. 7, 2017, 9:42 a.m. UTC | #2
On Fri, Jan 06, 2017 at 01:36:11PM -0800, John Stultz wrote:
> On Thu, Jan 5, 2017 at 6:50 PM, Leo Yan <leo.yan@linaro.org> wrote:

> > When use configfs to configure USB port as as ethernet gadget, the

> > kernel has panic with below backtrace; it clearly indicates the

> > coherent memory allocation is happened in the interrupt context,

> > but the function __get_vm_area_node() is possible to be scheduling

> > out so function __get_vm_area_node() triggers panic if it's called

> > from interrupt context.

> >

> > To fix this issue, simply to change gfp_flag from GFP_KERNEL to

> > GFP_NOWAIT so it can support to allocate coherent memory from

> > interrupt context.

> 

> This looks like it duplicates the similar patch I'm told should be

> queued for 4.10-rc3:

>     https://lkml.org/lkml/2016/12/1/120


Yeah, this patch is duplicate. Please ignore this one.

Thanks,
Leo Yan
diff mbox

Patch

diff --git a/drivers/usb/dwc2/gadget.c b/drivers/usb/dwc2/gadget.c
index b95930f..302b60c 100644
--- a/drivers/usb/dwc2/gadget.c
+++ b/drivers/usb/dwc2/gadget.c
@@ -3753,7 +3753,7 @@  static int dwc2_hsotg_ep_enable(struct usb_ep *ep,
 		hs_ep->desc_list = dma_alloc_coherent(hsotg->dev,
 			MAX_DMA_DESC_NUM_GENERIC *
 			sizeof(struct dwc2_dma_desc),
-			&hs_ep->desc_list_dma, GFP_KERNEL);
+			&hs_ep->desc_list_dma, GFP_NOWAIT);
 		if (!hs_ep->desc_list) {
 			ret = -ENOMEM;
 			goto error2;