On 9/11/2020 12:18 AM, Greg KH wrote:
Thus the TPM get queried more frequently than before, which is likely the root cause of the atmel chip crash. We fix it by bumping up the TPM_TIMEOUT to 15ms.
Rubrik Patch
```
diff --git a/drivers/char/tpm/tpm.h b/drivers/char/tpm/tpm.h
index 72d3ce4..9b8f3f8 100644
--- a/drivers/char/tpm/tpm.h
+++ b/drivers/char/tpm/tpm.h
@@ -49,7 +49,15 @@ enum tpm_const {
};
enum tpm_timeout {
- TPM_TIMEOUT = 5, /* msecs */
+ TPM_TIMEOUT = 15, /* msecs */
TPM_TIMEOUT_RETRY = 100, /* msecs */
TPM_TIMEOUT_RANGE_US = 300 /* usecs */
};
```
With the patch, the atmel TPM chip crash is fixed.
Proposal
We want the kernel upstream to adopt the fix or have a better fix for the atmel chip while not bring performance regression for other TPM chips. We understand thathttps://github.com/torvalds/linux/commit/9f3fc7bcddcb51234e23494531f93ab60475e1c3 <https://github.com/torvalds/linux/commit/9f3fc7bcddcb51234e23494531f93ab60475e1c3> was intended to shorten the TPM
Is this the poll time, which was reduced at one point? If so ...
Be careful about making this a global change. It could reduce the TPM
performance by 3x. We don't want to affect all TPMs to fix a bug in an
old TPM 1.2 chip from one vendor.
