On 14.09.2021 23:55, Bjorn Helgaas wrote:
> On Tue, Sep 14, 2021 at 07:07:40PM +0200, Heiner Kallweit wrote:
>> On 14.09.2021 13:26, Bjorn Helgaas wrote:
>>> On Tue, Sep 14, 2021 at 08:21:46AM +0200, Heiner Kallweit wrote:
>>>> On 14.09.2021 01:46, Bjorn Helgaas wrote:
>>>
>>>>> +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_ATTANSIC, PCI_ANY_ID, quirk_blacklist_vpd);
>>>>> /*
>>>>
>>>> Leaving the quirks in FIXUP_HEADER stage would have the advantage that for
>>>> blacklisted devices the vpd sysfs attribute isn't visibale. The needed
>>>> changes to the patch are minimal.
>>>
>>> What do you have in mind? The only thing I can think of would be to
>>> add a "pci_dev.no_vpd" bit. "vpd.cap == 0" means the device has no
>>> VPD, and "vpd.len == 0" means we haven't determined the size yet. All
>>> devices start off with vpd.cap == 0 and vpd.len == 0, so a
>>> FIXUP_HEADER quirk would have to set a sentinel value or some other
>>> bit.
>>
>> Why not leave vpd.len == PCI_VPD_SZ_INVALID as sentinel?
>
> Sentinel values aren't really my favorite thing, but it certainly does
> have the advantage of hiding the sysfs attribute.
>
>> And one more question: Why do you move the "if (!vpd->cap)" check from
>> pci_vpd_read() to pci_read_vpd()? At a first glance I see no benefit.
>
> I'm pretty sure I *had* a reason, but I can't remember right now :(
> Moving it sure does uglify pci_read_vpd() and pci_write_vpd(), though.
>
> What do you think of the following? (This is a diff from v5.15-rc1.)
>
> diff --git a/drivers/pci/vpd.c b/drivers/pci/vpd.c
> index 25557b272a4f..4be24890132e 100644
> --- a/drivers/pci/vpd.c
> +++ b/drivers/pci/vpd.c
> @@ -99,6 +99,24 @@ static size_t pci_vpd_size(struct pci_dev *dev)
> return off ?: PCI_VPD_SZ_INVALID;
> }
>
> +static bool pci_vpd_available(struct pci_dev *dev)
> +{
> + struct pci_vpd *vpd = &dev->vpd;
> +
> + if (!vpd->cap)
> + return false;
> +
> + if (vpd->len == 0) {
> + vpd->len = pci_vpd_size(dev);
> + if (vpd->len == PCI_VPD_SZ_INVALID) {
> + vpd->cap = 0;
> + return false;
> + }
> + }
> +
> + return true;
> +}
> +
> /*
> * Wait for last operation to complete.
> * This code has to spin since there is no other notification from the PCI
> @@ -145,7 +163,7 @@ static ssize_t pci_vpd_read(struct pci_dev *dev, loff_t pos, size_t count,
> loff_t end = pos + count;
> u8 *buf = arg;
>
> - if (!vpd->cap)
> + if (!pci_vpd_available(dev))
> return -ENODEV;
>
> if (pos < 0)
> @@ -206,7 +224,7 @@ static ssize_t pci_vpd_write(struct pci_dev *dev, loff_t pos, size_t count,
> loff_t end = pos + count;
> int ret = 0;
>
> - if (!vpd->cap)
> + if (!pci_vpd_available(dev))
> return -ENODEV;
>
> if (pos < 0 || (pos & 3) || (count & 3))
> @@ -242,14 +260,11 @@ static ssize_t pci_vpd_write(struct pci_dev *dev, loff_t pos, size_t count,
>
> void pci_vpd_init(struct pci_dev *dev)
> {
> + if (dev->vpd.len == PCI_VPD_SZ_INVALID)
> + return;
> +
> dev->vpd.cap = pci_find_capability(dev, PCI_CAP_ID_VPD);
> mutex_init(&dev->vpd.lock);
> -
> - if (!dev->vpd.len)
> - dev->vpd.len = pci_vpd_size(dev);
> -
> - if (dev->vpd.len == PCI_VPD_SZ_INVALID)
> - dev->vpd.cap = 0;
> }
>
> static ssize_t vpd_read(struct file *filp, struct kobject *kobj,
> @@ -294,13 +309,14 @@ const struct attribute_group pci_dev_vpd_attr_group = {
>
> void *pci_vpd_alloc(struct pci_dev *dev, unsigned int *size)
> {
> - unsigned int len = dev->vpd.len;
> + unsigned int len;
> void *buf;
> int cnt;
>
> - if (!dev->vpd.cap)
> + if (!pci_vpd_available(dev))
> return ERR_PTR(-ENODEV);
>
> + len = dev->vpd.len;
> buf = kmalloc(len, GFP_KERNEL);
> if (!buf)
> return ERR_PTR(-ENOMEM);
>
This looks very good to me.
