On Mon, 2017-01-16 at 19:24 +0200, Jarkko Sakkinen wrote: > On Mon, Jan 16, 2017 at 09:14:13AM -0700, Jason Gunthorpe wrote: > > On Mon, Jan 16, 2017 at 03:12:11PM +0200, Jarkko Sakkinen wrote: > > > > > @@ -199,7 +227,9 @@ struct tpm_chip *tpm_chip_alloc(struct device > > > *pdev, > > > return chip; > > > > > > out: > > > + put_device(&chip->devrm); > > > put_device(&chip->dev); > > > + put_device(&chip->devrm); > > > return ERR_PTR(rc); > > > } > > > > Something has gone wrong here.. > > Not a big surprise. There were a bunch of these small patches > that I had to squash. Thanks for spotting this. > > Funny that I didn't experiece any issues when I run my smoke > tests.
It's the error path in tpm_chip_alloc: it's incredibly hard to trigger, which is probably why your tests didn't see it. You basically either have to be out of kernel memory or out of TPM device numbers. James
