* [PATCH] nvme-cli: Wait for device file if not present after successful add_ctrl
@ 2018-04-27 17:29 James Smart
2018-04-27 21:45 ` Keith Busch
0 siblings, 1 reply; 2+ messages in thread
From: James Smart @ 2018-04-27 17:29 UTC (permalink / raw)
It's possible for the transport to return very quickly after add_ctrl
such that the cli may attempt to access the /dev/nvme? device file
before the udev event has propagated to user space to create the device
file. In these cases, the open fails with EAGAIN.
As the add_ctrl call was successful, thus there "should" be a device
file, if the open fails with EAGAIN delay 500ms and try again. Wait
for at most 2 seconds for the device file to come into existence.
Signed-off-by: James Smart <james.smart at broadcom.com>
---
fabrics.c | 28 ++++++++++++++++++++++++++++
1 file changed, 28 insertions(+)
diff --git a/fabrics.c b/fabrics.c
index 7e9d56c..408afb0 100644
--- a/fabrics.c
+++ b/fabrics.c
@@ -171,6 +171,32 @@ static const char *cms_str(__u8 cm)
static int do_discover(char *argstr, bool connect);
+#define OPEN_RETRIES 4
+#define OPEN_DELAY 500000 /* us units, thus 1/2 second */
+
+static void
+wait_for_devname(int instance)
+{
+ char *dev_name;
+ int retry = 0, fd;
+
+ if (asprintf(&dev_name, "/dev/nvme%d", instance) < 0)
+ return;
+
+ /* takes a little time for udev to make the dev node */
+ for ( ; retry < OPEN_RETRIES; retry++) {
+ fd = open(dev_name, O_RDWR);
+ if (fd >= 0) {
+ close(fd);
+ break;
+ }
+ if (errno != EAGAIN)
+ break;
+ usleep(OPEN_DELAY);
+ }
+ free(dev_name);
+}
+
static int add_ctrl(const char *argstr)
{
substring_t args[MAX_OPT_ARGS];
@@ -225,6 +251,8 @@ out_fail:
out_close:
close(fd);
out:
+ if (ret >= 0)
+ wait_for_devname(ret);
return ret;
}
--
2.13.1
^ permalink raw reply related [flat|nested] 2+ messages in thread
* [PATCH] nvme-cli: Wait for device file if not present after successful add_ctrl
2018-04-27 17:29 [PATCH] nvme-cli: Wait for device file if not present after successful add_ctrl James Smart
@ 2018-04-27 21:45 ` Keith Busch
0 siblings, 0 replies; 2+ messages in thread
From: Keith Busch @ 2018-04-27 21:45 UTC (permalink / raw)
On Fri, Apr 27, 2018@10:29:39AM -0700, James Smart wrote:
> It's possible for the transport to return very quickly after add_ctrl
> such that the cli may attempt to access the /dev/nvme? device file
> before the udev event has propagated to user space to create the device
> file. In these cases, the open fails with EAGAIN.
>
> As the add_ctrl call was successful, thus there "should" be a device
> file, if the open fails with EAGAIN delay 500ms and try again. Wait
> for at most 2 seconds for the device file to come into existence.
>
> Signed-off-by: James Smart <james.smart at broadcom.com>
Applied.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2018-04-27 21:45 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-04-27 17:29 [PATCH] nvme-cli: Wait for device file if not present after successful add_ctrl James Smart
2018-04-27 21:45 ` Keith Busch
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.