20 Apr
2023
20 Apr
'23
7:25 a.m.
On Wed, Apr 19, 2023 at 06:34:09PM -0700, Jakub Kicinski wrote:
On Wed, 19 Apr 2023 15:37:57 +0000 David Laight wrote:
Before the referenced commit, if fewer interrupts are supported by hardware than requested, then pci_msix_vec_count() returned the former. However, after the referenced commit, an error is returned for this condition. This causes a regression in the NFP driver preventing probe from completing.
I believe the relevant change to the msix vector allocation function has been reverted. (Or at least, the over-zealous check of nvec removed.)
So this change to bound the number of interrupts isn't needed.
Great, thanks, I was about to ask!
Likewise, thanks. We'll look into this.