Re: [PATCH v7 02/26] PCI: Enable bridge's I/O and MEM access for hotplugged devices

From: Bjorn Helgaas <helgaas@kernel.org>
To: Sergei Miroshnichenko <s.miroshnichenko@yadro.com>
Cc: linux-pci@vger.kernel.org, Stefan Roese <sr@denx.de>, linux@yadro.com
Subject: Re: [PATCH v7 02/26] PCI: Enable bridge's I/O and MEM access for hotplugged devices
Date: Thu, 30 Jan 2020 17:12:22 -0600	[thread overview]
Message-ID: <20200130231222.GA141544@google.com> (raw)
In-Reply-To: <20200129152937.311162-3-s.miroshnichenko@yadro.com>

On Wed, Jan 29, 2020 at 06:29:13PM +0300, Sergei Miroshnichenko wrote:
> The PCI_COMMAND_IO and PCI_COMMAND_MEMORY bits of the bridge must be
> updated not only when enabling the bridge for the first time, but also if a
> hotplugged device requests these types of resources.
> 
> For example, if a bridge had all its slots empty, the IO and MEM bits will
> not be set, and a hot hotplugged device will fail.

s/hot hotplugged/hot-added/ or something similar

> Originally these bits were set by the pci_enable_device_flags() only, which
> exits early if the bridge is already pci_is_enabled(). So let's check them
> again when requested.

s/by the/by/
s/ only,/,

IIUC, in the current tree (before this series), we do set
PCI_COMMAND_IO and PCI_COMMAND_MEMORY on bridges leading to hot-added
devices, but this patch is needed because [01/26] "PCI: Fix race
condition in pci_enable/disable_device()" makes
pci_enable_device_flags() exit early without setting those bits.

That would mean there's a bisection hole between [01/26] and [02/26]
where hot-added devices will fail.  We don't want a hole like that.

> Signed-off-by: Sergei Miroshnichenko <s.miroshnichenko@yadro.com>
> ---
>  drivers/pci/pci.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c
> index 0366289c75e9..cb0042f28e6a 100644
> --- a/drivers/pci/pci.c
> +++ b/drivers/pci/pci.c
> @@ -1679,6 +1679,14 @@ static void pci_enable_bridge(struct pci_dev *dev)
>  		pci_enable_bridge(bridge);
>  
>  	if (pci_is_enabled(dev)) {
> +		int i, bars = 0;
> +
> +		for (i = PCI_BRIDGE_RESOURCES; i < DEVICE_COUNT_RESOURCE; i++) {
> +			if (dev->resource[i].flags & (IORESOURCE_MEM | IORESOURCE_IO))
> +				bars |= (1 << i);
> +		}
> +		do_pci_enable_device(dev, bars);
> +
>  		if (!dev->is_busmaster)
>  			pci_set_master(dev);
>  		mutex_unlock(&dev->enable_mutex);
> -- 
> 2.24.1
> 