Hi,
> > Ok, I think this isn't needed anymore with the latest 5.15.116
> > release, right? If not, please let me know.
> >
>
>
> With 0c7e66e5fd ("tpm, tpm_tis: Request threaded interrupt handler") applied
> the above bug is fixed in 5.15.y. There is however still the issue that the interrupts
> may not be acknowledged properly in the interrupt handler, since the concerning
> register is written without the required locality held (Felix mentions this above).
> This can be fixed with 0e069265bce5 ("tpm, tpm_tis: Claim locality in interrupt
> handler").
>
> So instead of reverting the initial patches, I suggest to
>
> 1. also apply 0e069265bce5 ("tpm, tpm_tis: Claim locality in interrupt handler")
>
Yes, that one's needed to avoid the IRQ being permanently triggered.
If I remember correctly ( I disabled the IRQ for now) e644b2f498 ("tpm,
tpm_tis: Enable interrupt test") was also necessary to fix the interrupt
testing logic.
Without it the interrupt wouldn't be tested and the driver would poll the TPM
again. But since the test didn't execute, the IRQ would also stay enabled in
both the TPM and the host. So, looking at /proc/interrupts one could see the
TPM's IRQ counter being registered and counting up during use.
That's where I stopped trying because I was unsure if the remaining patches
from Lino's patchset were also necessary. It didn't do anything obviously bad
when I poked the TPM for random numbers or hashes for a quick test. But I
didn't test other (more complex) TPM operations then.
Regards,
Felix
___________________________________________________
SMA Solar Technology AG
Vorsitzender des Aufsichtsrats: Uwe Kleinkauf
Vorstand: Dr.-Ing. Juergen Reinert (Vorstandssprecher), Barbara Gregor
Handelsregister: Amtsgericht Kassel HRB 3972
Sitz der Gesellschaft: 34266 Niestetal
USt-ID-Nr. DE 113 08 59 54
WEEE-Reg.-Nr. DE 95881150
___________________________________________________
