[dpdk-dev,V11,2/3] eal: add uevent pass and process function

Message ID 1516013331-18939-2-git-send-email-jia.guo@intel.com (mailing list archive)
State Superseded, archived
Delegated to: Thomas Monjalon
Headers

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation fail Compilation issues

Commit Message

Guo, Jia Jan. 15, 2018, 10:48 a.m. UTC
In order to handle the uevent which have been detected from the kernel
side, add uevent process function, let hot plug event to be example to
show uevent mechanism how to pass the uevent and process the uevent.

About uevent passing and processing, add below functions in linux eal
dev layer. FreeBSD not support uevent ,so let it to be void and do not
implement in function.
a.dev_uev_parse
b.dev_uev_receive
c.dev_uev_process

Signed-off-by: Jeff Guo <jia.guo@intel.com>
---
v11->v10:
fix some typo issue.
---
 lib/librte_eal/common/include/rte_dev.h |  17 +++++
 lib/librte_eal/linuxapp/eal/eal_dev.c   | 110 +++++++++++++++++++++++++++++++-
 2 files changed, 125 insertions(+), 2 deletions(-)
  

Comments

Thomas Monjalon Jan. 17, 2018, 10 p.m. UTC | #1
15/01/2018 11:48, Jeff Guo:
> +enum rte_dev_event_subsystem {
> +       RTE_DEV_EVENT_SUBSYSTEM_UIO,
> +       RTE_DEV_EVENT_SUBSYSTEM_VFIO,
> +       RTE_DEV_EVENT_SUBSYSTEM_PCI,
> +       RTE_DEV_EVENT_SUBSYSTEM_MAX
> +};

I still don't understand this classification, mixing PCI and VFIO
at the same level.
  
Guo, Jia Jan. 18, 2018, 4:17 a.m. UTC | #2
On 1/18/2018 6:00 AM, Thomas Monjalon wrote:
> 15/01/2018 11:48, Jeff Guo:
>> +enum rte_dev_event_subsystem {
>> +       RTE_DEV_EVENT_SUBSYSTEM_UIO,
>> +       RTE_DEV_EVENT_SUBSYSTEM_VFIO,
>> +       RTE_DEV_EVENT_SUBSYSTEM_PCI,
>> +       RTE_DEV_EVENT_SUBSYSTEM_MAX
>> +};
> I still don't understand this classification, mixing PCI and VFIO
> at the same level.
ok, so let me explicit explain that that is because the deference kernel 
version would poke deference subsystem info from the net link message. 
so  just forcus on the new kernel version would be fine and delete pci item.
  

Patch

diff --git a/lib/librte_eal/common/include/rte_dev.h b/lib/librte_eal/common/include/rte_dev.h
index f6c9acb..0dbbaa8 100644
--- a/lib/librte_eal/common/include/rte_dev.h
+++ b/lib/librte_eal/common/include/rte_dev.h
@@ -51,6 +51,23 @@  extern "C" {
 
 #include <rte_log.h>
 
+#define RTE_EAL_UEV_MSG_LEN 4096
+#define RTE_EAL_UEV_MSG_ELEM_LEN 128
+
+enum rte_dev_state {
+	RTE_DEV_UNDEFINED,	/**< unknown device state */
+	RTE_DEV_FAULT,	/**< device fault or error */
+	RTE_DEV_PARSED,	/**< device has been scanned on bus*/
+	RTE_DEV_PROBED,	/**< device has been probed driver  */
+};
+
+enum rte_dev_event_subsystem {
+	RTE_DEV_EVENT_SUBSYSTEM_UIO,
+	RTE_DEV_EVENT_SUBSYSTEM_VFIO,
+	RTE_DEV_EVENT_SUBSYSTEM_PCI,
+	RTE_DEV_EVENT_SUBSYSTEM_MAX
+};
+
 /**
  * The device event type.
  */
diff --git a/lib/librte_eal/linuxapp/eal/eal_dev.c b/lib/librte_eal/linuxapp/eal/eal_dev.c
index 4f4beb5..bda4618 100644
--- a/lib/librte_eal/linuxapp/eal/eal_dev.c
+++ b/lib/librte_eal/linuxapp/eal/eal_dev.c
@@ -79,10 +79,116 @@  dev_monitor_enable(int netlink_fd)
 	return -1;
 }
 
+static void
+dev_uev_parse(const char *buf, struct rte_dev_event *event)
+{
+	char action[RTE_EAL_UEV_MSG_ELEM_LEN];
+	char subsystem[RTE_EAL_UEV_MSG_ELEM_LEN];
+	char dev_path[RTE_EAL_UEV_MSG_ELEM_LEN];
+	char pci_slot_name[RTE_EAL_UEV_MSG_ELEM_LEN];
+	int i = 0;
+
+	memset(action, 0, RTE_EAL_UEV_MSG_ELEM_LEN);
+	memset(subsystem, 0, RTE_EAL_UEV_MSG_ELEM_LEN);
+	memset(dev_path, 0, RTE_EAL_UEV_MSG_ELEM_LEN);
+	memset(pci_slot_name, 0, RTE_EAL_UEV_MSG_ELEM_LEN);
+
+	while (i < RTE_EAL_UEV_MSG_LEN) {
+		for (; i < RTE_EAL_UEV_MSG_LEN; i++) {
+			if (*buf)
+				break;
+			buf++;
+		}
+		if (!strncmp(buf, "libudev", 7)) {
+			buf += 7;
+			i += 7;
+			return;
+		}
+		if (!strncmp(buf, "ACTION=", 7)) {
+			buf += 7;
+			i += 7;
+			snprintf(action, sizeof(action), "%s", buf);
+		} else if (!strncmp(buf, "DEVPATH=", 8)) {
+			buf += 8;
+			i += 8;
+			snprintf(dev_path, sizeof(dev_path), "%s", buf);
+		} else if (!strncmp(buf, "SUBSYSTEM=", 10)) {
+			buf += 10;
+			i += 10;
+			snprintf(subsystem, sizeof(subsystem), "%s", buf);
+		} else if (!strncmp(buf, "PCI_SLOT_NAME=", 14)) {
+			buf += 14;
+			i += 14;
+			snprintf(pci_slot_name, sizeof(subsystem), "%s", buf);
+			event->devname = pci_slot_name;
+		}
+		for (; i < RTE_EAL_UEV_MSG_LEN; i++) {
+			if (*buf == '\0')
+				break;
+			buf++;
+		}
+	}
+
+	if (!strncmp(subsystem, "pci", 3))
+		event->subsystem = RTE_DEV_EVENT_SUBSYSTEM_UIO;
+	if (!strncmp(action, "add", 3))
+		event->type = RTE_DEV_EVENT_ADD;
+	if (!strncmp(action, "remove", 6))
+		event->type = RTE_DEV_EVENT_REMOVE;
+	event->devname = pci_slot_name;
+}
+
+static int
+dev_uev_receive(int fd, struct rte_dev_event *uevent)
+{
+	int ret;
+	char buf[RTE_EAL_UEV_MSG_LEN];
+
+	memset(uevent, 0, sizeof(struct rte_dev_event));
+	memset(buf, 0, RTE_EAL_UEV_MSG_LEN);
+
+	ret = recv(fd, buf, RTE_EAL_UEV_MSG_LEN - 1, MSG_DONTWAIT);
+	if (ret < 0) {
+		RTE_LOG(ERR, EAL,
+		"Socket read error(%d): %s\n",
+		errno, strerror(errno));
+		return -1;
+	} else if (ret == 0)
+		/* connection closed */
+		return -1;
+
+	dev_uev_parse(buf, uevent);
+
+	return 0;
+}
+
 static int
-dev_uev_process(__rte_unused struct epoll_event *events, __rte_unused int nfds)
+dev_uev_process(struct epoll_event *events, int nfds)
 {
-	/* TODO: device uevent processing */
+	struct rte_dev_event uevent;
+	int i;
+
+	for (i = 0; i < nfds; i++) {
+		/**
+		 * check device uevent from kernel side, no need to check
+		 * uevent from udev.
+		 */
+		if (dev_uev_receive(events[i].data.fd, &uevent))
+			return 0;
+
+		/* default handle all pci devcie when is being hot plug */
+		if (uevent.subsystem == RTE_DEV_EVENT_SUBSYSTEM_UIO) {
+			if (uevent.type == RTE_DEV_EVENT_REMOVE) {
+				return(_rte_dev_callback_process(
+					uevent.devname,
+					RTE_DEV_EVENT_REMOVE, NULL));
+			} else if (uevent.type == RTE_DEV_EVENT_ADD) {
+				return(_rte_dev_callback_process(
+					uevent.devname,
+					RTE_DEV_EVENT_ADD, NULL));
+			}
+		}
+	}
 	return 0;
 }