The tpm2_get_cc_attrs_tbl() call will result in TPM commands being issued,
which will need the use of the internal command/response buffer.  But,
we're issuing this *before* we've waited to make sure that buffer is
allocated.
This can result in intermittent failures to probe if the hypervisor / TPM
implementation doesn't respond quickly enough.  I find it fails almost
every time with an 8 vcpu guest under KVM with software emulated TPM.
To fix it, just move the tpm2_get_cc_attrs_tlb() call after the
existing code to wait for initialization, which will ensure the buffer
is allocated.
Fixes: 18b3670d79ae9 ("tpm: ibmvtpm: Add support for TPM2")
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Jerry Snitselaar <jsnitsel@redhat.com>
Reviewed-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
        if (rc)
                goto init_irq_cleanup;
 
-       if (!strcmp(id->compat, "IBM,vtpm20")) {
-               chip->flags |= TPM_CHIP_FLAG_TPM2;
-               rc = tpm2_get_cc_attrs_tbl(chip);
-               if (rc)
-                       goto init_irq_cleanup;
-       }
-
        if (!wait_event_timeout(ibmvtpm->crq_queue.wq,
                                ibmvtpm->rtce_buf != NULL,
                                HZ)) {
                goto init_irq_cleanup;
        }
 
+       if (!strcmp(id->compat, "IBM,vtpm20")) {
+               chip->flags |= TPM_CHIP_FLAG_TPM2;
+               rc = tpm2_get_cc_attrs_tbl(chip);
+               if (rc)
+                       goto init_irq_cleanup;
+       }
+
        return tpm_chip_register(chip);
 init_irq_cleanup:
        do {