2009-07-27 22:17:21

by Daniel J Blueman

Subject: [2.6.31-rc4] ftdi_usb maps memory from stack...

With DMA debugging enabled, I see a warning when the FTDI USB-serial
driver loads [1] using a non-DMAable address.

I can get a disassembly of some of the stack frames to identify the
exact call-sites etc, if that helps anyone?

Thanks,
Daniel

--- [1]

usbcore: registered new interface driver usbserial_generic

usbserial: USB Serial Driver core

USB Serial support registered for FTDI USB Serial Device

ftdi_sio 5-1:1.0: FTDI USB Serial Device converter detected

usb 5-1: Detected FT8U232AM

usb 5-1: Number of endpoints 2

usb 5-1: Endpoint 1 MaxPacketSize 64

usb 5-1: Endpoint 2 MaxPacketSize 64

usb 5-1: Setting MaxPacketSize 64

------------[ cut here ]------------

WARNING: at lib/dma-debug.c:856 check_for_stack+0x9a/0xc0()

Hardware name: OEM

uhci_hcd 0000:00:1d.0: DMA-API: device driver maps memory from stack
[addr=ffff88007f89b946]

Modules linked in:

Pid: 1, comm: swapper Not tainted 2.6.31-rc4-274sd #1

Call Trace:

[<ffffffff8124796a>] ? check_for_stack+0x9a/0xc0

[<ffffffff810490f8>] warn_slowpath_common+0x78/0xd0

[<ffffffff810491d4>] warn_slowpath_fmt+0x64/0x70

[<ffffffff81430379>] ? error_exit+0x29/0xb0

[<ffffffff812afa61>] ? dev_driver_string+0x1/0x40

[<ffffffff8124796a>] check_for_stack+0x9a/0xc0

[<ffffffff812490ff>] debug_dma_map_page+0xef/0x160

[<ffffffff81314e21>] usb_hcd_submit_urb+0x361/0xc90

[<ffffffff8142f23d>] ? trace_hardirqs_off_thunk+0x3a/0x3c

[<ffffffff81430379>] ? error_exit+0x29/0xb0

[<ffffffff81315b65>] usb_submit_urb+0xd5/0x260

[<ffffffff81317403>] usb_start_wait_urb+0x63/0xf0

[<ffffffff81317720>] usb_control_msg+0x130/0x160

[<ffffffff81339c2a>] ftdi_sio_port_probe+0x3ea/0x7f0

[<ffffffff81336255>] usb_serial_device_probe+0x75/0x110

[<ffffffff812b3ad5>] driver_probe_device+0x95/0x1a0

[<ffffffff812b3c80>] ? __device_attach+0x0/0x50

[<ffffffff812b3ccb>] __device_attach+0x4b/0x50

[<ffffffff812b2ebb>] bus_for_each_drv+0x6b/0xa0

[<ffffffff812b3d98>] device_attach+0x88/0x90

[<ffffffff812b2c95>] bus_attach_device+0x55/0x80

[<ffffffff812b126a>] device_add+0x42a/0x670

[<ffffffff812b14c9>] device_register+0x19/0x20

[<ffffffff813340f4>] usb_serial_probe+0x1274/0x1550

[<ffffffff8102f992>] ? kmemcheck_fault+0x72/0x90

[<ffffffff8142f23d>] ? trace_hardirqs_off_thunk+0x3a/0x3c

[<ffffffff81430379>] ? error_exit+0x29/0xb0

[<ffffffff8131a182>] usb_probe_interface+0xc2/0x1a0

[<ffffffff812b3ad5>] driver_probe_device+0x95/0x1a0

[<ffffffff812b3c73>] __driver_attach+0x93/0xa0

[<ffffffff812b3be0>] ? __driver_attach+0x0/0xa0

[<ffffffff812b325b>] bus_for_each_dev+0x6b/0xa0

[<ffffffff812b392c>] driver_attach+0x1c/0x20

[<ffffffff812b2ae8>] bus_add_driver+0x248/0x2b0

[<ffffffff812b3f7c>] driver_register+0x7c/0x170

[<ffffffff81319ede>] usb_register_driver+0xce/0x140

[<ffffffff81676660>] ? ftdi_init+0x0/0xe2

[<ffffffff8167671e>] ftdi_init+0xbe/0xe2

[<ffffffff8100905b>] do_one_initcall+0x4b/0x1a0

[<ffffffff81654525>] kernel_init+0x1b2/0x236

[<ffffffff8100ce9a>] child_rip+0xa/0x20

[<ffffffff8100c83c>] ? restore_args+0x0/0x30

[<ffffffff81654373>] ? kernel_init+0x0/0x236

[<ffffffff8100ce90>] ? child_rip+0x0/0x20

---[ end trace 8ecb5c2eb0ae54d6 ]---

usb 5-1: FTDI USB Serial Device converter now attached to ttyUSB0

usbcore: registered new interface driver ftdi_sio

ftdi_sio: v1.5.0:USB FTDI Serial Converters Driver

--
Daniel J Blueman

2009-07-28 19:31:31

by Robert Hancock

Subject: Re: [2.6.31-rc4] ftdi_usb maps memory from stack...

On 07/27/2009 04:17 PM, Daniel J Blueman wrote:
> With DMA debugging enabled, I see a warning when the FTDI USB-serial
> driver loads [1] using a non-DMAable address.
>
> I can get a disassembly of some of the stack frames to identify the
> exact call-sites etc, if that helps anyone?
>
> Thanks,
> Daniel
>
> --- [1]
>
> usbcore: registered new interface driver usbserial_generic
>
> usbserial: USB Serial Driver core
>
> USB Serial support registered for FTDI USB Serial Device
>
> ftdi_sio 5-1:1.0: FTDI USB Serial Device converter detected
>
> usb 5-1: Detected FT8U232AM
>
> usb 5-1: Number of endpoints 2
>
> usb 5-1: Endpoint 1 MaxPacketSize 64
>
> usb 5-1: Endpoint 2 MaxPacketSize 64
>
> usb 5-1: Setting MaxPacketSize 64
>
> ------------[ cut here ]------------
>
> WARNING: at lib/dma-debug.c:856 check_for_stack+0x9a/0xc0()
>
> Hardware name: OEM
>
> uhci_hcd 0000:00:1d.0: DMA-API: device driver maps memory from stack
> [addr=ffff88007f89b946]

Looks like some of the functions in ftdi_sio.c call usb_control_msg with
a stack buffer and some use a kmalloc buffer. They should likely all be
using a kmalloc buffer.