Linux PCI subsystem development
 help / color / mirror / Atom feed
From: Bjorn Helgaas <helgaas@kernel.org>
To: Ethan Zhao <haifeng.zhao@intel.com>
Cc: bhelgaas@google.com, oohall@gmail.com, ruscur@russell.cc,
	lukas@wunner.de, andriy.shevchenko@linux.intel.com,
	stuart.w.hayes@gmail.com, mr.nuke.me@gmail.com,
	mika.westerberg@linux.intel.com, linux-pci@vger.kernel.org,
	linux-kernel@vger.kernel.org, ashok.raj@linux.intel.com,
	sathyanarayanan.kuppuswamy@intel.com, xerces.zhao@gmail.com,
	Sinan Kaya <okaya@kernel.org>
Subject: Re: [PATCH v6 4/5] PCI: only return true when dev io state is really changed
Date: Fri, 2 Oct 2020 12:29:13 -0500	[thread overview]
Message-ID: <20201002172913.GA2809822@bjorn-Precision-5520> (raw)
In-Reply-To: <20200930070537.30982-5-haifeng.zhao@intel.com>

[+cc Sinan]

On Wed, Sep 30, 2020 at 03:05:36AM -0400, Ethan Zhao wrote:
> When uncorrectable error happens, AER driver and DPC driver interrupt
> handlers likely call
> 
>    pcie_do_recovery()
>    ->pci_walk_bus()
>      ->report_frozen_detected()
> 
> with pci_channel_io_frozen the same time.
>    If pci_dev_set_io_state() return true even if the original state is
> pci_channel_io_frozen, that will cause AER or DPC handler re-enter
> the error detecting and recovery procedure one after another.
>    The result is the recovery flow mixed between AER and DPC.
> So simplify the pci_dev_set_io_state() function to only return true
> when dev->error_state is changed.
> 
> Signed-off-by: Ethan Zhao <haifeng.zhao@intel.com>
> Tested-by: Wen Jin <wen.jin@intel.com>
> Tested-by: Shanshan Zhang <ShanshanX.Zhang@intel.com>
> Reviewed-by: Alexandru Gagniuc <mr.nuke.me@gmail.com>
> Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
> ---
> Changnes:
>  v2: revise description and code according to suggestion from Andy.
>  v3: change code to simpler.
>  v4: no change.
>  v5: no change.
>  v6: no change.
> 
>  drivers/pci/pci.h | 37 +++++--------------------------------
>  1 file changed, 5 insertions(+), 32 deletions(-)
> 
> diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
> index 455b32187abd..f2beeaeda321 100644
> --- a/drivers/pci/pci.h
> +++ b/drivers/pci/pci.h
> @@ -359,39 +359,12 @@ struct pci_sriov {
>  static inline bool pci_dev_set_io_state(struct pci_dev *dev,
>  					pci_channel_state_t new)
>  {
> -	bool changed = false;
> -
>  	device_lock_assert(&dev->dev);
> -	switch (new) {
> -	case pci_channel_io_perm_failure:
> -		switch (dev->error_state) {
> -		case pci_channel_io_frozen:
> -		case pci_channel_io_normal:
> -		case pci_channel_io_perm_failure:
> -			changed = true;
> -			break;
> -		}
> -		break;
> -	case pci_channel_io_frozen:
> -		switch (dev->error_state) {
> -		case pci_channel_io_frozen:
> -		case pci_channel_io_normal:
> -			changed = true;
> -			break;
> -		}
> -		break;
> -	case pci_channel_io_normal:
> -		switch (dev->error_state) {
> -		case pci_channel_io_frozen:
> -		case pci_channel_io_normal:
> -			changed = true;
> -			break;
> -		}
> -		break;
> -	}
> -	if (changed)
> -		dev->error_state = new;
> -	return changed;
> +	if (dev->error_state == new)
> +		return false;
> +
> +	dev->error_state = new;
> +	return true;
>  }

IIUC this changes the behavior of the function, but it's difficult to
analyze because it does a lot of simplification at the same time.

Please consider the following, which is intended to simplify the
function while preserving the behavior (but please verify; it's been a
long time since I looked at this).  Then maybe see how your patch
could be done on top of this?

Alternatively, come up with your own simplification patch + the
functionality change.


commit 983d9b1f8177 ("PCI/ERR: Simplify pci_dev_set_io_state()")
Author: Bjorn Helgaas <bhelgaas@google.com>
Date:   Tue May 19 12:28:57 2020 -0500

    PCI/ERR: Simplify pci_dev_set_io_state()
    
    Truth table:
    
                                  requested new state
      current          ------------------------------------------
      state            normal         frozen         perm_failure
      ------------  +  -------------  -------------  ------------
      normal        |  normal         frozen         perm_failure
      frozen        |  normal         frozen         perm_failure
      perm_failure  |  perm_failure*  perm_failure*  perm_failure
    
      * "not changed", returns false
    
    No functional change intended.
    
diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
index 6d3f75867106..81408552f7c9 100644
--- a/drivers/pci/pci.h
+++ b/drivers/pci/pci.h
@@ -358,39 +358,21 @@ struct pci_sriov {
 static inline bool pci_dev_set_io_state(struct pci_dev *dev,
 					pci_channel_state_t new)
 {
-	bool changed = false;
-
 	device_lock_assert(&dev->dev);
-	switch (new) {
-	case pci_channel_io_perm_failure:
-		switch (dev->error_state) {
-		case pci_channel_io_frozen:
-		case pci_channel_io_normal:
-		case pci_channel_io_perm_failure:
-			changed = true;
-			break;
-		}
-		break;
-	case pci_channel_io_frozen:
-		switch (dev->error_state) {
-		case pci_channel_io_frozen:
-		case pci_channel_io_normal:
-			changed = true;
-			break;
-		}
-		break;
-	case pci_channel_io_normal:
-		switch (dev->error_state) {
-		case pci_channel_io_frozen:
-		case pci_channel_io_normal:
-			changed = true;
-			break;
-		}
-		break;
+
+	/* Can always put a device in perm_failure state */
+	if (new == pci_channel_io_perm_failure) {
+		dev->error_state == pci_channel_io_perm_failure;
+		return true;
 	}
-	if (changed)
-		dev->error_state = new;
-	return changed;
+
+	/* If already in perm_failure, can't set to normal or frozen */
+	if (dev->error_state == pci_channel_io_perm_failure)
+		return false;
+
+	/* Can always change normal to frozen or vice versa */
+	dev->error_state = new;
+	return true;
 }
 
 static inline int pci_dev_set_disconnected(struct pci_dev *dev, void *unused)

  parent reply	other threads:[~2020-10-02 17:29 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-09-30  7:05 [PATCH v6 0/5] Fix DPC hotplug race and enhance error handling Ethan Zhao
2020-09-30  7:05 ` [PATCH v6 1/5] PCI/ERR: get device before call device driver to avoid NULL pointer dereference Ethan Zhao
2020-09-30  7:05 ` [PATCH v6 2/5] PCI/DPC: define a function to check and wait till port finish DPC handling Ethan Zhao
2020-09-30  7:05 ` [PATCH v6 3/5] PCI: pciehp: check and wait port status out of DPC before handling DLLSC and PDC Ethan Zhao
2020-09-30  7:05 ` [PATCH v6 4/5] PCI: only return true when dev io state is really changed Ethan Zhao
2020-10-02 16:08   ` Sinan Kaya
2020-10-03  5:46     ` Ethan Zhao
2020-10-02 17:29   ` Bjorn Helgaas [this message]
2020-10-03  7:05     ` Ethan Zhao
2020-09-30  7:05 ` [PATCH v6 5/5] PCI/ERR: don't mix io state not changed and no driver together Ethan Zhao

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20201002172913.GA2809822@bjorn-Precision-5520 \
    --to=helgaas@kernel.org \
    --cc=andriy.shevchenko@linux.intel.com \
    --cc=ashok.raj@linux.intel.com \
    --cc=bhelgaas@google.com \
    --cc=haifeng.zhao@intel.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-pci@vger.kernel.org \
    --cc=lukas@wunner.de \
    --cc=mika.westerberg@linux.intel.com \
    --cc=mr.nuke.me@gmail.com \
    --cc=okaya@kernel.org \
    --cc=oohall@gmail.com \
    --cc=ruscur@russell.cc \
    --cc=sathyanarayanan.kuppuswamy@intel.com \
    --cc=stuart.w.hayes@gmail.com \
    --cc=xerces.zhao@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox