Re: [PATCH] net: fec: properly guard irq coalesce setup

From: Rasmus Villemoes
Date: Tue Dec 06 2022 - 03:23:45 EST


On 06/12/2022 07.16, Alexander Stein wrote:
> Am Montag, 5. Dezember 2022, 21:46:04 CET schrieb Rasmus Villemoes:
>> Prior to the Fixes: commit, the initialization code went through the
>> same fec_enet_set_coalesce() function as used by ethtool, and that
>> function correctly checks whether the current variant has support for
>> irq coalescing.
>>
>> Now that the initialization code instead calls fec_enet_itr_coal_set()
>> directly, that call needs to be guarded by a check for the
>> FEC_QUIRK_HAS_COALESCE bit.
>>
>> Fixes: df727d4547de (net: fec: don't reset irq coalesce settings to defaults
>> on "ip link up") Reported-by: Greg Ungerer <gregungerer@xxxxxxxxxxxxxx>
>> Signed-off-by: Rasmus Villemoes <linux@xxxxxxxxxxxxxxxxxx>
>> ---
>> drivers/net/ethernet/freescale/fec_main.c | 3 ++-
>> 1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/ethernet/freescale/fec_main.c
>> b/drivers/net/ethernet/freescale/fec_main.c index
>> 2ca2b61b451f..23e1a94b9ce4 100644
>> --- a/drivers/net/ethernet/freescale/fec_main.c
>> +++ b/drivers/net/ethernet/freescale/fec_main.c
>> @@ -1220,7 +1220,8 @@ fec_restart(struct net_device *ndev)
>> writel(0, fep->hwp + FEC_IMASK);
>>
>> /* Init the interrupt coalescing */
>> - fec_enet_itr_coal_set(ndev);
>> + if (fep->quirks & FEC_QUIRK_HAS_COALESCE)
>> + fec_enet_itr_coal_set(ndev);
>> }
>>
>> static int fec_enet_ipc_handle_init(struct fec_enet_private *fep)
>
> I'm wondering if this check should be added to fec_enet_itr_coal_set()
> instead. Right now any additional caller has to do it's own check for
> FEC_QUIRK_HAS_COALESCE, so why not do check in fec_enet_itr_coal_set?

Yeah, I did consider that (see
https://lore.kernel.org/netdev/72eb4e63-10a8-702b-1113-7588fcade9fc@xxxxxxxxxxxxxxxxxx/).
But in the end I decided against it, for various small reasons.
Primarily my thinking was that while this caller is obviously ok with
fec_enet_itr_coal_set() just silently being a no-op, it's not clear that
some future caller wouldn't have to check the bit themselves and return
-EOPNOTSUPP or some other error just like the ethtool method does. Of
course we could also make fec_enet_itr_coal_set() not return void, but
then it's a whole different refactoring and definitely not for post-rc8.

But also, the above was what I had sent out as a "something like this?"
and tested on my own hardware before heading home, and due to the
imminent release, I wanted to send out the patch ASAP.

Rasmus