Re: [PATCH] [net/irda]: new Blackfin on-chip SIR IrDA driver

From: graff yang
Date: Thu Mar 12 2009 - 03:44:10 EST


On Thu, Mar 12, 2009 at 2:24 PM, Mike Frysinger <vapier.adi@xxxxxxxxx> wrote:
> On Thu, Mar 12, 2009 at 01:55, graff yang wrote:
>> On Thu, Mar 12, 2009 at 12:59 PM, Mike Frysinger wrote:
>>> On Thu, Mar 12, 2009 at 00:48, gyang wrote:
>>>> On Thu, 2009-03-12 at 00:34 -0400, Mike Frysinger wrote:
>>>>> On Thu, Mar 12, 2009 at 00:30, gyang wrote:
>>>>> > On Thu, 2009-03-12 at 00:23 -0400, Mike Frysinger wrote:
>>>>> >> On Thu, Mar 12, 2009 at 00:17, gyang wrote:
>>>>> >> > On Wed, 2009-03-11 at 06:43 -0400, Mike Frysinger wrote:
>>>>> >> >> On Wed, Mar 11, 2009 at 05:56, graff yang wrote:
>>>>> >> >> > On Wed, Mar 11, 2009 at 3:57 PM, Mike Frysinger wrote:
>>>>> >> >> >> On Wed, Mar 11, 2009 at 03:29, Â<graff.yang@xxxxxxxxx> wrote:
>>>>> >> >> >>> +static int __devinit bfin_sir_probe(struct platform_device *pdev)
>>>>> >> >> >>> +{
>>>>> >> >> >>> + Â Â Â struct net_device *dev;
>>>>> >> >> >>> + Â Â Â struct bfin_sir_self *self;
>>>>> >> >> >>> + Â Â Â unsigned int baudrate_mask;
>>>>> >> >> >>> + Â Â Â struct bfin_sir_port *sir_port;
>>>>> >> >> >>> + Â Â Â int err;
>>>>> >> >> >>> +
>>>>> >> >> >>> + Â Â Â err = peripheral_request_list(per[pdev->id], DRIVER_NAME);
>>>>> >> >> >>
>>>>> >> >> >> what if pdev->id is set to 12512 ?
>>>>> >> >> >
>>>>> >> >> > The pdev->id is defined in board files, for example, uart0 should be 0.
>>>>> >> >>
>>>>> >> >> so ? Âwhat's to stop the user from setting it to 12415 ? Âthe driver
>>>>> >> >> must sanity check these things. Âyou can never assume platform dev
>>>>> >> >> resources always have valid pointers and valid values.
>>>>> >> >
>>>>> >> > How about to limit the pdev->id to be 0,1,2,3 ?
>>>>> >> > I can extend the per to be
>>>>> >> > static const unsigned short per[][4] = {
>>>>> >> > Â Â Â Â{P_UART0_RX, P_UART0_TX, 0, 0},
>>>>> >> > Â Â Â Â{P_UART1_RX, P_UART1_TX, 0, 1},
>>>>> >> > Â Â Â Â{P_UART2_RX, P_UART2_TX, 0, 2},
>>>>> >> > Â Â Â Â{P_UART3_RX, P_UART3_TX, 0, 3},
>>>>> >> > };
>>>>> >> >
>>>>> >> > then check (pdev->id >= 0 && pdev->id < ARRAY_SIZE(per) &&
>>>>> >> > per[pdev->id][3] == pdev->id)
>>>>> >>
>>>>> >> the last check is redundant. Âthe first two should be added.
>>>>> >
>>>>> > I'm afraid of some cases such as uart0 uses id 1, will cause driver not
>>>>> > work, and not any warning message.
>>>>>
>>>>> then the call to peripheral request would fail as expected
>>>>
>>>> For example, user want to enable irda on uart2 for bf548, he use id 0
>>>> for uart2. In these case, peripheral request may not fail, but driver
>>>> not work.
>>>
>>> the id is wrong then. Âif they want uart2, then they should use id 2.
>>
>> Such wrong case may existed for common users, and should be
>> prohibited. Driver need make sanity check to prevent it.
>
> i dont know what you're talking about. Âif the platform resource uses
> id of 2, then they want uart2 which means irda2.

The X of irdaX is determined by the order of calling register_netdev().
It will be irda0 for the first registered irda device, in spite of the pdev->id.

Because you argued that user can use any id for pdev->id, I think I not only
need limit them to be 0~3, but also need keep them in right order, or else,
driver may be fail.
It is easy for driver to check that, and safe for users.

> the sanity check is
> already in place because the peripheral_request() will fail when it
> attempts to claim the pins of uart2 but the uart driver has already
> been enabled for it.
> -mike
>



--
-Graff
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/