On 10/8/25 12:51, Akihiko Odaki wrote:
On 2025/10/07 2:19, Michael Tokarev wrote:
On 9/1/25 18:14, CLEMENT MATHIEU--DRIF wrote:
From: Damien Bergamini <[email protected]>
Starting with commit cab1398a60eb, SR-IOV VFs are realized as soon as
pcie_sriov_pf_init() is called. Because pcie_sriov_pf_init() must be
called before pcie_sriov_pf_init_vf_bar(), the VF BARs types won't be
known when the VF realize function calls pcie_sriov_vf_register_bar().
This breaks the memory regions of the VFs (for instance with igbvf):
...
Fixes: cab1398a60eb ("pcie_sriov: Reuse SR-IOV VF device instances")
Signed-off-by: Damien Bergamini <[email protected]>
Signed-off-by: Clement Mathieu--Drif <[email protected]>
This one too, - is it qemu-stable material (10.0 & 10.1)?
Yes, I think so.
I haven't picked it up for the tagged-today stable releases
because of this:
For 10.0.x (which is a long-term support series), it needs
some adjustments I guess (it doesn't apply to 10.0 directly).
For 10.0 (which is an long-term series), this change clashes with
19e55471d4 "pcie_sriov: Allow user to create SR-IOV device", which
is not in 10.0.
Is it worth to try to back-port it to 10.0.x? If yes, I'd need
some help with doing so :)
Thank you!
/mjt
