MTL OFI: Allow retries in MTL progress for interrupted syscalls

aravindksg · aravindksg · commit 0822f571618a · 2017-12-21T11:08:06.000-08:00
This fixes a regression in sockets provider which could return -EINTR value from fi_cq_read() due to a syscall being interrupted. The error value is currently interpreted as fatal condition. Relax the rule so that we can retry fi_cq_read() operation. Signed-off-by: Aravind Gopalakrishnan <Aravind.Gopalakrishnan@intel.com> (cherry picked from commit fb68726)
diff --git a/ompi/mca/mtl/ofi/mtl_ofi.h b/ompi/mca/mtl/ofi/mtl_ofi.h
@@ -116,7 +116,7 @@ ompi_mtl_ofi_progress(void)
                 exit(1);
             }
         } else {
-            if (ret == -FI_EAGAIN) {
+            if (ret == -FI_EAGAIN || ret == -EINTR) {
                 break;
             } else {
                 opal_output(0, "%s:%d: Error returned from fi_cq_read: %s(%zd).\n"

Original file line number	Diff line number	Diff line change
`@@ -116,7 +116,7 @@ ompi_mtl_ofi_progress(void)`
`116`	`116`	`exit(1);`
`117`	`117`	`}`
`118`	`118`	`} else {`
`119`		`- if (ret == -FI_EAGAIN) {`
	`119`	`+ if (ret == -FI_EAGAIN \|\| ret == -EINTR) {`
`120`	`120`	`break;`
`121`	`121`	`} else {`
`122`	`122`	`opal_output(0, "%s:%d: Error returned from fi_cq_read: %s(%zd).\n"`