Re: [PATCH] igc: Return early when failing to read EECD register

[Lifshits, Vitaly]

On 12/17/2024 3:23 AM, Chia-Lin Kao (AceLan) wrote:
> On Mon, Dec 16, 2024 at 06:53:10AM +0100, Przemek Kitszel wrote:
> > On 12/16/24 06:14, Chia-Lin Kao (AceLan) wrote:
> > > When booting with a dock connected, the igc driver can get stuck for ~40
> > > seconds if PCIe link is lost during initialization.
> > >
> > > This happens because the driver access device after EECD register reads
> > > return all F's, indicating failed reads. Consequently, hw->hw_addr is set
> > > to NULL, which impacts subsequent rd32() reads. This leads to the driver
> > > hanging in igc_get_hw_semaphore_i225(), as the invalid hw->hw_addr
> > > prevents retrieving the expected value.
> >
> > Than you very much for the patch and the analysis!
> >
> > > To address this, a validation check is added for the EECD register read
> > > result. If all F's are returned, indicating PCIe link loss, the driver
> > > will return -ENXIO immediately. This avoids the 40-second hang and
> >
> > It is not clear from the patch what part of the driver will return
> > -ENXIO, you have put -ENODEV in the patch, but it's ignored anyway.
> I was thinking of using -ENODEV or -ENXIO, and I forgot to generate
> the patch again after I changed it to -ENXIO in the commit message.
> I'll fix this in v2.
> > > significantly improves boot time when using a dock with an igc NIC.
> > >
> > > [    0.911913] igc 0000:70:00.0: enabling device (0000 -> 0002)
> > > [    0.912386] igc 0000:70:00.0: PTM enabled, 4ns granularity
> > > [    1.571098] igc 0000:70:00.0 (unnamed net_device) (uninitialized): PCIe link lost, device now detached
> > > [   43.449095] igc_get_hw_semaphore_i225: igc 0000:70:00.0 (unnamed net_device) (uninitialized): Driver can't access device - SMBI bit is set.
> > > [   43.449186] igc 0000:70:00.0: probe with driver igc failed with error -13
> > > [   46.345701] igc 0000:70:00.0: enabling device (0000 -> 0002)
> > > [   46.345777] igc 0000:70:00.0: PTM enabled, 4ns granularity
> >
> > Would be best if you could also attach the sequence after your fix.
> Sure
>
> > Please add a Fixes: tag.
> I'm not sure which commit should I add Fixes to.
>
> > Please make [PATCH iwl-net] as a subject prefix. Please CC Vitaly.
> igc is an ethernet driver, should I also add iwl-net tag?
>
> > (But please also wait a day prior to sending v2 for more feedback).
> >
> > > Signed-off-by: Chia-Lin Kao (AceLan) <acelan.kao@xxxxxxxxxxxxx>
> > > ---
> > >    drivers/net/ethernet/intel/igc/igc_base.c | 4 ++++
> > >    1 file changed, 4 insertions(+)
> > >
> > > diff --git a/drivers/net/ethernet/intel/igc/igc_base.c b/drivers/net/ethernet/intel/igc/igc_base.c
> > > index 9fae8bdec2a7..54ce60280765 100644
> > > --- a/drivers/net/ethernet/intel/igc/igc_base.c
> > > +++ b/drivers/net/ethernet/intel/igc/igc_base.c
> > > @@ -68,6 +68,10 @@ static s32 igc_init_nvm_params_base(struct igc_hw *hw)
> >
> > This function is used only in igc_get_invariants_base(), which ignores
> > the return value you have added. I would expect it to propagate instead.
> You are right, looks like the patch fixes the issue accidentally.
> Return earlier in igc_get_invariants_base() skipping the rest of the
> settings. The part impacts the behavior is nvm->word_size will be 0.
> And then in igc_get_hw_semaphore_i225() the timeout value will become
> 1. So that we won't hang for 40 seconds to wait for the timeout.
>
> This patch is not perfect, I need to figure out another way to address
> the issue better.
> Please let me know if you got any ideas.
> Thanks.


I like your approach to this bug fix.

However, it seems that without another change in igc_get_invariants_base 
the failure won't cause an early return with an error status.
You will need to add something like:
	ret_val = igc_init_nvm_params_base(hw);
+       if (ret_val)
+		return -ENODEV;
	switch (hw->mac.type) {


> > >    	u32 eecd = rd32(IGC_EECD);
> > >    	u16 size;
> > > +	/* failed to read reg and got all F's */
> > > +	if (!(~eecd))
> > > +		return -ENODEV;
> > > +
> > >    	size = FIELD_GET(IGC_EECD_SIZE_EX_MASK, eecd);
> > >    	/* Added to a constant, "size" becomes the left-shift value