Re: [PATCH v2] PCI: Reprogram bridge prefetch registers on resume

[Peter Wu <peter@lekensteyn.nl>]

On Wed, Sep 12, 2018 at 02:45:23PM +0800, Daniel Drake wrote:
> On 38+ Intel-based Asus products, the nvidia GPU becomes unusable
> after S3 suspend/resume. The affected products include multiple
> generations of nvidia GPUs and Intel SoCs. After resume, nouveau logs
> many errors such as:
> 
>     fifo: fault 00 [READ] at 0000005555555000 engine 00 [GR] client 04 [HUB/FE] reason 4a [] on channel -1 [007fa91000 unknown]
>     DRM: failed to idle channel 0 [DRM]
> 
> Similarly, the nvidia proprietary driver also fails after resume
> (black screen, 100% CPU usage in Xorg process). We shipped a sample
> to Nvidia for diagnosis, and their response indicated that it's a
> problem with the parent PCI bridge (on the Intel SoC), not the GPU.
> 
> Runtime suspend/resume works fine, only S3 suspend is affected.
> 
> We found a workaround: on resume, rewrite the Intel PCI bridge
> 'Prefetchable Base Upper 32 Bits' register (PCI_PREF_BASE_UPPER32). In
> the cases that I checked, this register has value 0 and we just have to
> rewrite that value.
> 
> Linux already saves and restores PCI config space during suspend/resume,
> but this register was being skipped because upon resume, it already
> has value 0 (the correct, pre-suspend value).
> 
> Intel appear to have previously acknowledged this behaviour and the
> requirement to rewrite this register.
> https://bugzilla.kernel.org/show_bug.cgi?id=116851#c23
> 
> Based on that, rewrite the prefetch register values even when that
> appears unnecessary.
> 
> We have confirmed this solution on all the affected models we have
> in-hands (X542UQ, UX533FD, X530UN, V272UN).
> 
> Additionally, this solves an issue where r8169 MSI-X interrupts were
> broken after S3 suspend/resume on Asus X441UAR. This issue was recently
> worked around in commit 7bb05b85bc2d ("r8169: don't use MSI-X on
> RTL8106e"). It also fixes the same issue on RTL6186evl/8111evl on an
> Aimfor-tech laptop that we had not yet patched. I suspect it will also
> fix the issue that was worked around in commit 7c53a722459c ("r8169:
> don't use MSI-X on RTL8168g").
> 
> Thomas Martitz reports that this change also solves an issue where
> the AMD Radeon Polaris 10 GPU on the HP Zbook 14u G5 is unresponsive
> after S3 suspend/resume.
> 
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=201069
> Signed-off-by: Daniel Drake <drake@endlessm.com>
> ---
> 
> Notes:
>     Replaces patch:
>        PCI: add prefetch quirk to work around Asus/Nvidia suspend issues
>     
>     Some of the more verbose info was moved to bugzilla:
>     https://bugzilla.kernel.org/show_bug.cgi?id=201069
>     
>     This patch is aimed at v4.19 (and maybe v4.18-stable); we may follow
>     up with more intrusive improvements for v4.20+.
>     
>     v2: reimplement the register restore within the existing
>         pci_restore_config_space() code.
> 
>  drivers/pci/pci.c | 31 +++++++++++++++++++++----------
>  1 file changed, 21 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 29ff9619b5fa..e1704100e72d 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -1289,13 +1289,15 @@ int pci_save_state(struct pci_dev *dev)
>  EXPORT_SYMBOL(pci_save_state);
>  
>  static void pci_restore_config_dword(struct pci_dev *pdev, int offset,
> -				     u32 saved_val, int retry)
> +				     u32 saved_val, int retry, bool force)
>  {
>  	u32 val;
>  
> -	pci_read_config_dword(pdev, offset, &val);
> -	if (val == saved_val)
> -		return;
> +	if (!force) {
> +		pci_read_config_dword(pdev, offset, &val);
> +		if (val == saved_val)
> +			return;
> +	}
>  
>  	for (;;) {
>  		pci_dbg(pdev, "restoring config space at offset %#x (was %#x, writing %#x)\n",

"val" is uninitialized and used in this debug message.

To avoid adding another parameter and changing unrelated lines below,
what about:

    force = pdev->header_type == PCI_HEADER_TYPE_BRIDGE &&
        0x24 <= start && start <= 0x2c;

(or start == 0x28 || start == 0x24 || start == 0x28 if you want to be
very explicit). This would reduce in a smaller diff for this temporary
solution.

Kind regards,
Peter

> @@ -1313,25 +1315,34 @@ static void pci_restore_config_dword(struct pci_dev *pdev, int offset,
>  }
>  
>  static void pci_restore_config_space_range(struct pci_dev *pdev,
> -					   int start, int end, int retry)
> +					   int start, int end, int retry,
> +					   bool force)
>  {
>  	int index;
>  
>  	for (index = end; index >= start; index--)
>  		pci_restore_config_dword(pdev, 4 * index,
>  					 pdev->saved_config_space[index],
> -					 retry);
> +					 retry, force);
>  }
>  
>  static void pci_restore_config_space(struct pci_dev *pdev)
>  {
>  	if (pdev->hdr_type == PCI_HEADER_TYPE_NORMAL) {
> -		pci_restore_config_space_range(pdev, 10, 15, 0);
> +		pci_restore_config_space_range(pdev, 10, 15, 0, false);
>  		/* Restore BARs before the command register. */
> -		pci_restore_config_space_range(pdev, 4, 9, 10);
> -		pci_restore_config_space_range(pdev, 0, 3, 0);
> +		pci_restore_config_space_range(pdev, 4, 9, 10, false);
> +		pci_restore_config_space_range(pdev, 0, 3, 0, false);
> +	} else if (pdev->hdr_type == PCI_HEADER_TYPE_BRIDGE) {
> +		pci_restore_config_space_range(pdev, 12, 15, 0, false);
> +		/* Force rewriting of prefetch registers to avoid
> +		 * S3 resume issues on Intel PCI bridges that occur when
> +		 * these registers are not explicitly written.
> +		 */
> +		pci_restore_config_space_range(pdev, 9, 11, 0, true);
> +		pci_restore_config_space_range(pdev, 0, 8, 0, false);
>  	} else {
> -		pci_restore_config_space_range(pdev, 0, 15, 0);
> +		pci_restore_config_space_range(pdev, 0, 15, 0, false);
>  	}
>  }
>  
> -- 
> 2.17.1
>