[<prev] [next>] [day] [month] [year] [list]
Message-ID: <c5d1b5d8-8669-5572-75a7-0b480f581ac1@linux.intel.com>
Date: Tue, 12 Nov 2024 16:28:55 +0200 (EET)
From: Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com>
To: Christian König <christian.koenig@....com>
cc: Michał Winiarski <michal.winiarski@...el.com>,
Bjorn Helgaas <bhelgaas@...gle.com>, linux-pci@...r.kernel.org,
LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] PCI: Fix BAR resizing when VF BARs are assigned
On Tue, 12 Nov 2024, Christian König wrote:
> Am 12.11.24 um 14:42 schrieb Ilpo Järvinen:
>
> __resource_resize_store() attempts to release all resources of the
> device before attempting the resize. The loop, however, only covers
> standard BARs (< PCI_STD_NUM_BARS). If a device has VF BARs that are
> assigned, pci_reassign_bridge_resources() finds the bridge window still
> has some assigned child resources and returns -NOENT which makes
> pci_resize_resource() to detect an error and abort the resize.
>
>
> Looks good in general, but a couple of comments.
>
> Similarly, an assigned Expansion ROM resource prevents the resize.
>
>
> Expansion ROMs were intentionally left untouched by the initial patch since those are
> usually 32bit resources and the resize was implemented only for 64bit BARs.
>
> Change the release loop to cover both the VF BARs and Expansion ROM
> which allows the resize operation to release the bridge windows and
> attempt to assigned them again with the different size.
>
>
> I'm not sure if Expansion ROMs should be touched here. Those are 32bit resources
> usually and notoriously broken in the ACPI tables.
>
> What I've seen multiple times that after releasing them you can't move nor assign
> them again to their original position.
>
> Background is that some ACPI table denotes the location of the ROM as reserved and we
> only accept the Expansion ROM at that location because of a quirk (which is of course
> not applied when you resize).
>
> On the other hand do we have use cases for resizing 32bit BARs? So far we resized
> those only by accident.
Thanks for taking a look!
This is not about resizing 32bit BARs.
Is that somehow enforced so they cannot end up into the same bridge
window? Because we can only resize a bridge window if we release all its
child resources.
> As __resource_resize_store() checks first that no driver is bound to
> the PCI device before resizing is allowed, SR-IOV cannot be enabled
> during resize so it is safe to release also the IOV resources.
>
> Fixes: 8bb705e3e79d ("PCI: Add pci_resize_resource() for resizing BARs")
>
>
> The code in question was added by 91fa127794ac ("PCI: Expose PCIe Resizable BAR
> support via sysfs").
Oh right. I don't know why I ended up picking that other commit (sure, I
was also looking that other diff but maybe it was just a copy-paste
error).
--
i.
> Regards,
> Christian.
>
> Reported-by: Michał Winiarski <michal.winiarski@...el.com>
> Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@...ux.intel.com>
> ---
> drivers/pci/pci-sysfs.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/pci/pci-sysfs.c b/drivers/pci/pci-sysfs.c
> index 5d0f4db1cab7..80b01087d3ef 100644
> --- a/drivers/pci/pci-sysfs.c
> +++ b/drivers/pci/pci-sysfs.c
> @@ -1440,7 +1440,7 @@ static ssize_t __resource_resize_store(struct device *dev, int
> n,
>
> pci_remove_resource_files(pdev);
>
> - for (i = 0; i < PCI_STD_NUM_BARS; i++) {
> + for (i = 0; i < PCI_BRIDGE_RESOURCES; i++) {
> if (pci_resource_len(pdev, i) &&
> pci_resource_flags(pdev, i) == flags)
> pci_release_resource(pdev, i);
>
>
>
>
Powered by blists - more mailing lists