On Mon, Nov 16, 2015 at 7:12 PM, Lars-Peter Clausen <l...@metafoo.de> wrote:
> Prior to commit e6c9a037bc8a ("i2c: xiic: Remove the disabling of
> interrupts") IRQs where disabled when the initial __xiic_start_xfer() was
> called. After the commit the interrupt is enabled while the function is
> running, this means it is possible for the interrupt to be triggered while
> the function is still running. When this happens the internal data
> structures get corrupted and undefined behavior can occur like the
> following crash:
>
>         Internal error: Oops: 17 [#1] PREEMPT SMP ARM
>         Modules linked in:
>         CPU: 0 PID: 2040 Comm: i2cdetect Not tainted 4.0.0-02856-g047a308 
> #10956
>         Hardware name: Xilinx Zynq Platform
>         task: ee0c9500 ti: e99a2000 task.ti: e99a2000
>         PC is at __xiic_start_xfer+0x6c4/0x7c8
>         LR is at __xiic_start_xfer+0x690/0x7c8
>         pc : [<c02bbffc>]    lr : [<c02bbfc8>]    psr: 800f0013
>         sp : e99a3da8  ip : 00000000  fp : 00000000
>         r10: 00000001  r9 : 600f0013  r8 : f0180000
>         r7 : f0180000  r6 : c064e444  r5 : 00000017  r4 : ee031010
>         r3 : 00000000  r2 : 00000000  r1 : 600f0013  r0 : 0000000f
>         Flags: Nzcv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment user
>         Control: 18c5387d  Table: 29a5404a  DAC: 00000015
>         Process i2cdetect (pid: 2040, stack limit = 0xe99a2210)
>         Stack: (0xe99a3da8 to 0xe99a4000)
>         3da0:                   ee031010 00000000 00000001 ee031020 ee031224 
> c02bc5ec
>         3dc0: ee34c604 00000000 ee0c9500 e99a3dcc e99a3dd0 e99a3dd0 e99a3dd8 
> c069f0e8
>         3de0: 00000000 ee031020 c064e100 ffff90bb e99a3e48 c02b6590 ee031020 
> 00000001
>         3e00: e99a3e48 ee031020 00000000 e99a3e63 00000001 c02b6ec4 00000000 
> 00000000
>         3e20: 00000000 c02b7320 e99a3ef0 00000000 00000000 e99e3df0 00000000 
> 00000000
>         3e40: 00000103 2814575f 0000003e c00a0000 e99a3e85 0001003e ee0c0000 
> e99a3e63
>         3e60: eefd3578 c064e61c ee0c9500 c0041e04 0000056c e9a56db8 00006e5a 
> b6f5c000
>         3e80: ee0c9548 eefd0040 00000001 eefd3540 ee0c9500 eefd39a0 c064b540 
> ee0c9500
>         3ea0: 00000000 ee92b000 00000000 bef4862c ee34c600 e99ecdc0 00000720 
> 00000003
>         3ec0: e99a2000 00000000 00000000 c02b8b30 00000000 00000000 00000000 
> e99a3f24
>         3ee0: b6e80000 00000000 00000000 c04257e8 00000000 e99a3f24 c02b8f08 
> 00000703
>         3f00: 00000003 c02116bc ee935300 00000000 bef4862c ee34c600 e99ecdc0 
> c02b91f0
>         3f20: e99ecdc0 00000720 bef4862c eeb725f8 e99ecdc0 c00c9e2c 00000003 
> 00000003
>         3f40: ee248dc0 00000000 ee248dc8 00000002 eeb7c1a8 00000000 00000000 
> c00bb360
>         3f60: 00000000 00000000 00000003 ee248dc0 bef4862c e99ecdc0 e99ecdc0 
> 00000720
>         3f80: 00000003 e99a2000 00000000 c00c9f68 00000000 00000000 b6f22000 
> 00000036
>         3fa0: c000dfa4 c000de20 00000000 00000000 00000003 00000720 bef4862c 
> bef4862c
>         3fc0: 00000000 00000000 b6f22000 00000036 00000000 00000000 b6f60000 
> 00000000
>         3fe0: 00013040 bef48614 00008cab b6ecdbe6 400f0030 00000003 2f7fd821 
> 2f7fdc21
>         [<c02bbffc>] (__xiic_start_xfer) from [<c02bc5ec>] 
> (xiic_xfer+0x94/0x168)
>         [<c02bc5ec>] (xiic_xfer) from [<c02b6590>] (__i2c_transfer+0x4c/0x7c)
>         [<c02b6590>] (__i2c_transfer) from [<c02b6ec4>] 
> (i2c_transfer+0x9c/0xc4)
>         [<c02b6ec4>] (i2c_transfer) from [<c02b7320>] 
> (i2c_smbus_xfer+0x3a0/0x4ec)
>         [<c02b7320>] (i2c_smbus_xfer) from [<c02b8b30>] 
> (i2cdev_ioctl_smbus+0xb0/0x214)
>         [<c02b8b30>] (i2cdev_ioctl_smbus) from [<c02b91f0>] 
> (i2cdev_ioctl+0xa0/0x1d4)
>         [<c02b91f0>] (i2cdev_ioctl) from [<c00c9e2c>] 
> (do_vfs_ioctl+0x4b0/0x5b8)
>         [<c00c9e2c>] (do_vfs_ioctl) from [<c00c9f68>] (SyS_ioctl+0x34/0x5c)
>         [<c00c9f68>] (SyS_ioctl) from [<c000de20>] (ret_fast_syscall+0x0/0x34)
>         Code: e283300c e5843210 eafffe64 e5943210 (e1d320b4)
>
> The issue can easily be reproduced by performing I2C access under high
> system load or IO load.
>
> To fix the issue protect the invocation to __xiic_start_xfer() form
> xiic_start_xfer() with the same lock that is used to protect the interrupt
> handler.
>

Reviewed-by: Shubhrajyoti Datta <shubh...@xilinx.com>

Though I feel that this is the correct thing to do missed out in the
original code and exposed
by removing the disabling of interrupts. As disabling interrupts to
synchronize is crude.


> Fixes: e6c9a037bc8a ("i2c: xiic: Remove the disabling of interrupts")
> Signed-off-by: Lars-Peter Clausen <l...@metafoo.de>
> ---
>  drivers/i2c/busses/i2c-xiic.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/drivers/i2c/busses/i2c-xiic.c b/drivers/i2c/busses/i2c-xiic.c
> index 705cf69..0b20449 100644
> --- a/drivers/i2c/busses/i2c-xiic.c
> +++ b/drivers/i2c/busses/i2c-xiic.c
> @@ -664,9 +664,8 @@ static void xiic_start_xfer(struct xiic_i2c *i2c)
>  {
>         spin_lock(&i2c->lock);
>         xiic_reinit(i2c);
> -       spin_unlock(&i2c->lock);
> -
>         __xiic_start_xfer(i2c);
> +       spin_unlock(&i2c->lock);
>  }
>
>  static int xiic_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs, int num)
> --
> 2.1.4
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
> the body of a message to majord...@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to