RE: [PATCH v7 07/15] scsi: fnic: Add and integrate support for FDMI
From: Karan Tilak Kumar (kartilak)
Date: Mon Jan 13 2025 - 15:29:03 EST
On Monday, January 13, 2025 9:35 AM, John Meneghini <jmeneghi@xxxxxxxxxx> wrote:
>
> Just a note to say that these patches are important to Red Hat and we are actively engaged in back porting and testing these patches in to RHEL-9 and RHEL-10.
>
> I think these issues that Dan has pointed out are all issues which can be addressed in a follow up patch.
>
> /John
>
> On 1/7/25 07:30, Dan Carpenter wrote:
> > On Wed, Dec 11, 2024 at 06:03:04PM -0800, Karan Tilak Kumar wrote:
> >> @@ -612,6 +615,7 @@ static int fnic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
> >> unsigned long flags;
> >> int hwq;
> >> char *desc, *subsys_desc;
> >> + int len;
> >
> > Do not introduce unnecessary levels of indirection. Get rid of this len
> > variable.
> >
> >>
> >> /*
> >> * Allocate SCSI Host and set up association between host,
> >> @@ -646,9 +650,17 @@ static int fnic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
> >> fnic_stats_debugfs_init(fnic);
> >>
> >> /* Find model name from PCIe subsys ID */
> >> - if (fnic_get_desc_by_devid(pdev, &desc, &subsys_desc) == 0)
> >> + if (fnic_get_desc_by_devid(pdev, &desc, &subsys_desc) == 0) {
> >> dev_info(&fnic->pdev->dev, "Model: %s\n", subsys_desc);
> >> - else {
> >> +
> >> + /* Update FDMI model */
> >
> > This comment adds no information. Delete it.
> >
> >> + fnic->subsys_desc_len = strlen(subsys_desc);
> >
> > Keep in mind that strlen() does not count the NUL-terminator.
> >
> >> + len = ARRAY_SIZE(fnic->subsys_desc);
> >
> > Use sizeof() when you are talking about bytes or chars. For snprintf() and
> > other string functions, it's always sizeof() and never ARRAY_SIZE().
> >
> >> + if (fnic->subsys_desc_len > len)
> >> + fnic->subsys_desc_len = len;
> >> + memcpy(fnic->subsys_desc, subsys_desc, fnic->subsys_desc_len);
> >
> > So this is an 0-14 character buffer. If fnic->subsys_desc_len is set to 14,
> > then the string is not NUL terminated. This is how the buffer is used in
> > fdls_fdmi_register_hba()
> >
> > strscpy_pad(data, fnic->subsys_desc, FNIC_FDMI_MODEL_LEN);
> > data[FNIC_FDMI_MODEL_LEN - 1] = 0;
> >
> > This suggests that fnic->subsys_desc is expected to be NUL-terminated.
> > However FNIC_FDMI_MODEL_LEN is 12. So in that case the last 3 characters
> > are removed. LOL. It's harmless but so very annoying.
> >
> > Also strscpy_pad() will ensure that data[FNIC_FDMI_MODEL_LEN - 1] is set
> > to zero so that line could be deleted.
> >
> > regards,
> > dan carpenter
> >
>
>
Thanks for your inputs, John.
Regards,
Karan