Hi Sourabh,
Thanks for the review.
On 07/06/26 7:05 PM, Sourabh Jain wrote:
session_token 0 is considered valid here. Where as it was considered
invalid in other
function above.
Agreed. |rtas_open_errinjct_session()|is expected to return a positive
session token on
success, while zero is not considered a valid token by
|rtas_close_errinjct_session()|.
+ return session_token;
+
+ /* get errinjct token */
+ errinjct_token = rtas_function_token(RTAS_FN_IBM_ERRINJCT);
+ if (errinjct_token == RTAS_UNKNOWN_SERVICE) {
How about checking this before getting the session token?
Agreed. I will obtain and validate the |ibm,errinjct|RTAS token before
opening the error-injection session. This avoids opening a session when
the injection service is unavailable.
+ pr_err("RTAS: ibm,errinjct not available\n");
+ rc = -ENODEV;
+ goto out_close;
+ }
+
+ /* prepare shared buffer while holding lock */
+ spin_lock(&rtas_errinjct_buf_lock);
+ rc = prepare_errinjct_buffer(pe, type, func, addr, mask);
+ if (rc) {
+ spin_unlock(&rtas_errinjct_buf_lock);
+ goto out_close;
+ }
+
+ /* perform the errinjct RTAS call */
+ rc = do_errinjct_call(errinjct_token, type, session_token);
+ spin_unlock(&rtas_errinjct_buf_lock);
+
+out_close:
+ /* always attempt close if we opened a session */
+ rtas_close_errinjct_session(session_token);
+ return rc;
}
+
This new line seems unnecessary.
Agreed. I will remove the extra blank line.
Thanks,
Narayana
static struct eeh_ops pseries_eeh_ops = {
.name = "pseries",
.probe = pseries_eeh_probe,
