Re: [PATCH] PCI: dwc: Use the correct sleep function in wait_for_link

[Konrad Dybcio <konrad.dybcio@linaro.org>]

On 15.02.2024 18:02, Bjorn Helgaas wrote:
> On Thu, Feb 15, 2024 at 02:35:13PM +0100, Alexander Lobakin wrote:
>> From: Konrad Dybcio <konrad.dybcio@linaro.org>
>> Date: Thu, 15 Feb 2024 11:39:31 +0100
>>
>>> According to [1], msleep should be used for large sleeps, such as the
>>> 100-ish ms one in this function. Comply with the guide and use it.
>>>
>>> [1] https://www.kernel.org/doc/Documentation/timers/timers-howto.txt
>>>
>>> Signed-off-by: Konrad Dybcio <konrad.dybcio@linaro.org>
>>> ---
>>> Tested on Qualcomm SC8280XP CRD
>>> ---
>>>  drivers/pci/controller/dwc/pcie-designware.c | 2 +-
>>>  drivers/pci/controller/dwc/pcie-designware.h | 3 +--
>>>  2 files changed, 2 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/drivers/pci/controller/dwc/pcie-designware.c b/drivers/pci/controller/dwc/pcie-designware.c
>>> index 250cf7f40b85..abce6afceb91 100644
>>> --- a/drivers/pci/controller/dwc/pcie-designware.c
>>> +++ b/drivers/pci/controller/dwc/pcie-designware.c
>>> @@ -655,7 +655,7 @@ int dw_pcie_wait_for_link(struct dw_pcie *pci)
>>>  		if (dw_pcie_link_up(pci))
>>>  			break;
>>>  
>>> -		usleep_range(LINK_WAIT_USLEEP_MIN, LINK_WAIT_USLEEP_MAX);
>>> +		msleep(LINK_WAIT_MSLEEP_MAX);
>>
>> Just use fsleep(LINK_WAIT_USLEEP_MAX) and let the kernel decide which
>> function to pick.

IMO, fsleep only makes sense when the argument is variable.. This way, we
can save on bothering the compiler or adding an unnecessary branch

> 
> Odd.
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/timers/timers-howto.rst?id=v6.7#n114
> mentions fsleep() (with no real guidance about when to use it), but
> https://www.kernel.org/doc/Documentation/timers/timers-howto.txt 
> seems to be a stale copy from 2017, before fsleep() was added.  I
> emailed helpdesk@kernel.org to see if the stale content can be
> removed.
> 
> I do think fsleep() should be more widely used.
> 
>>>  /* Parameters for the waiting for link up routine */
>>>  #define LINK_WAIT_MAX_RETRIES		10
>>> -#define LINK_WAIT_USLEEP_MIN		90000
>>> -#define LINK_WAIT_USLEEP_MAX		100000
>>> +#define LINK_WAIT_MSLEEP_MAX		100
> 
> Since you're touching this anyway, it would be helpful to include the
> units on the timeout.
> 
> USLEEP/MSLEEP is definitely a hint, but I think the "SLEEP" part
> suggests something about atomic/non-atomic context and isn't relevant
> to the time interval itself, and something like "TIMEOUT" would be
> better.
> 
> I think an explicit "_US" or "_MS" would better indicate the units.
> 
> This is turning into a long tangent, but I'm not a huge fan of the
> LINK_WAIT_* pattern where I have to look up the code that uses
> LINK_WAIT_MAX_RETRIES and LINK_WAIT_USLEEP_MAX and do the math to see
> what the actual timeout is.  Obviously not fodder for *this* patch.

Might as well do that indeed

Konrad