On Fri, Apr 09, 2021 at 11:55:05AM +0200, Lino Sanfilippo wrote: > > Hi, > > > > > On 30/03/2021 03:04, Kees Cook wrote: > > > > > > Does this series solve the issue too? > > > > > > https://lore.kernel.org/linux-integrity/1613955394-13152-1-git-send-email-LinoSanfilippo@xxxxxx/ > > > > > > (I haven't had a chance to test either series with my TPM, but I see the > > > same "TPM returned invalid status" errors recently.) > > > > > > > Unfortunately no. I tested it immediately but forgot to let you know. > > > > > > kernel: Linux version 5.12.0-rc5tpm-fix+ (tj@elloe000) (gcc (Ubuntu > > 9.3.0-17ubuntu1~20.04) 9.3.0, GNU ld (GNU Binutils for Ubuntu) 2.34) #29 > > SMP PREEMPT Tue Mar 30 09:05:15 BST 2021 > > ... > > Well you tested the series with 5.12-rc5 while it was based on 5.11. Patch 2 > in the series is supposed to fix the "invalid status" warning by making sure > that the required locality has been requested before. > To avoid such issues at all and to simplify the whole locality handling (and > also to be prepared for interrupt handling which also requires locality management) > that patch furthermore ensures that the locality is requested only once at driver > startup and not released until driver shutdown. > > However between 5.11 and 5.12-rc5 there have been at least two patches that > again introduced a locality request/release combo (d53a6adfb553 "tpm, tpm_tis: > Decorate tpm_tis_gen_interrupt() with request_locality()" and a5665ec2affd > "tpm, tpm_tis: Decorate tpm_get_timeouts() with request_locality()"). > > The latter results in the locality being released again before tpm_tis_status() > is called and thus reintroduced the issue patch 2 fixed. > > I will prepare another series based on the latest kernel but at least for 5.11 > the series should fix the issue (and also make interrupts working). Thanks, I can review patches after coming from two week leave and include them to 5.13-rcX. Add stable cc in addition to the fixes tag. /Jarkko
