Re: [REGRESSION] tg3 is broken since 6.13-rc1

[Pavan Chebbi]

()

On Thu, Oct 29, 2026 at 5:42 PM Rui Salvaterra <rsalvaterra@xxxxxxxxx> wrote:
>
> Hi, Pavan,
>
>
> On Thu, 8 May 2025 at 10:14, Rui Salvaterra <rsalvaterra@xxxxxxxxx> wrote:
> >
> [snipped]
> >
> > > Can you please mention your model and the make again? Let me see if I
> > > can find one in our lab or near me.
> >
> > The machine is a late 2012 Mac Mini (DMI: Apple Inc. Macmini6,2).
> > From lspci -nn, the Ethernet controller is a...
> >
> > 01:00.0 Ethernet controller [0200]: Broadcom Inc. and subsidiaries
> > NetXtreme BCM57766 Gigabit Ethernet PCIe [14e4:1686] (rev 01)
>
> It's been over a year since I sent this email, and almost two years
> since I reported the regression. Is this ever going to be fixed, or do
> I have to carry the revert in my tree forever? From my dmesg, with
> Linux 7.3…
>

Hi Rui, first of all, apologies for the delay in providing you with a solution.
The only reason I did not pursue this report further was that I could
not explain why dma_set_coherent_mask() with lower mask would fail in
your system, while not being able to reproduce the error in our lab.
Now with the benefit of AI, I hope to figure that out.
In fact I fed your kconfig to it and looks like there is a probable
cause in CONFIG_ZONE_DMA not being set.
Let me investigate and get back asap..

> [    1.203519] tg3 0000:01:00.0: Unable to obtain 64 bit DMA for
> consistent allocations
> [    1.203584] tg3 0000:01:00.0: probe with driver tg3 failed with error -5
>
> My kconfig is attached.
>
>
> Kind regards,
>
> Rui Salvaterra

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature