Re: [PATCH 1/2] PCI: Correct error reporting with PCIe failed link retraining

[Bjorn Helgaas]

On Sat, Feb 10, 2024 at 01:43:50AM +0000, Maciej W. Rozycki wrote:
> Only return successful completion status from `pcie_failed_link_retrain' 
> if retraining has actually been done, preventing excessive delays from 
> being triggered at call sites in a hope that communication will finally 
> be established with the downstream device where in fact nothing has been 
> done about the link in question that would justify such a hope.
> 
> Fixes: a89c82249c37 ("PCI: Work around PCIe link training failures")
> Reported-by: Ilpo Järvinen <ilpo.jarvinen@xxxxxxxxxxxxxxx>
> Link: https://lore.kernel.org/r/aa2d1c4e-9961-d54a-00c7-ddf8e858a9b0@xxxxxxxxxxxxxxx/
> Signed-off-by: Maciej W. Rozycki <macro@xxxxxxxxxxx>
> Cc: stable@xxxxxxxxxxxxxxx # v6.5+
> ---
>  drivers/pci/quirks.c |   18 +++++++++++-------
>  1 file changed, 11 insertions(+), 7 deletions(-)
> 
> linux-pcie-failed-link-retrain-status-fix.diff
> Index: linux-macro/drivers/pci/quirks.c
> ===================================================================
> --- linux-macro.orig/drivers/pci/quirks.c
> +++ linux-macro/drivers/pci/quirks.c
> @@ -74,7 +74,8 @@
>   * firmware may have already arranged and lift it with ports that already
>   * report their data link being up.
>   *
> - * Return TRUE if the link has been successfully retrained, otherwise FALSE.
> + * Return TRUE if the link has been successfully retrained, otherwise FALSE,
> + * also when retraining was not needed in the first place.

Can you recast this?  I think it's slightly unclear what is returned
when retraining is not needed.  I *think* you return FALSE when
retraining is not needed.  Maybe this?

  Return TRUE if the link has been successfully retrained.  Return
  FALSE if retraining was not needed or we attempted a retrain and it
  failed.

>   */
>  bool pcie_failed_link_retrain(struct pci_dev *dev)
>  {
> @@ -83,10 +84,11 @@ bool pcie_failed_link_retrain(struct pci
>  		{}
>  	};
>  	u16 lnksta, lnkctl2;
> +	bool ret = false;
>  
>  	if (!pci_is_pcie(dev) || !pcie_downstream_port(dev) ||
>  	    !pcie_cap_has_lnkctl2(dev) || !dev->link_active_reporting)
> -		return false;
> +		return ret;

We know the value here, so IMO it's easier to read if we return
"false" instead of "ret".

>  	pcie_capability_read_word(dev, PCI_EXP_LNKCTL2, &lnkctl2);
>  	pcie_capability_read_word(dev, PCI_EXP_LNKSTA, &lnksta);
> @@ -98,9 +100,10 @@ bool pcie_failed_link_retrain(struct pci
>  		lnkctl2 |= PCI_EXP_LNKCTL2_TLS_2_5GT;
>  		pcie_capability_write_word(dev, PCI_EXP_LNKCTL2, lnkctl2);
>  
> -		if (pcie_retrain_link(dev, false)) {
> +		ret = pcie_retrain_link(dev, false) == 0;
> +		if (!ret) {
>  			pci_info(dev, "retraining failed\n");
> -			return false;
> +			return ret;

Same here.  We're here because !ret was true, so ret must be false.
I guess in the next patch you want to return the pcie_retrain_link()
return value up the chain, so it will make sense there.

>  		}
>  
>  		pcie_capability_read_word(dev, PCI_EXP_LNKSTA, &lnksta);
> @@ -117,13 +120,14 @@ bool pcie_failed_link_retrain(struct pci
>  		lnkctl2 |= lnkcap & PCI_EXP_LNKCAP_SLS;
>  		pcie_capability_write_word(dev, PCI_EXP_LNKCTL2, lnkctl2);
>  
> -		if (pcie_retrain_link(dev, false)) {
> +		ret = pcie_retrain_link(dev, false) == 0;
> +		if (!ret) {
>  			pci_info(dev, "retraining failed\n");
> -			return false;
> +			return ret;
>  		}
>  	}
>  
> -	return true;
> +	return ret;

It gets awfully subtle by the time we get here.  I guess we could set
a "retrain_attempted" flag above and do this:

  if (retrain_attempted)
    return true;

  return false;

But I dunno if it's any better.  I understand the need for a change
like this, but the whole idea of returning failure (false) for a
retrain failure and also for a "no retrain needed" is a little
mind-bending.

>  }
>  
>  static ktime_t fixup_debug_start(struct pci_dev *dev,