[PATCH 0/2] pci: allow to run without hotplug

[PATCH 0/2] pci: allow to run without hotplug

[Olivier Matz]

The default behavior of DPDK is to wait the creation of /dev/uioX
devices. This work is usually done by hotplug thanks to the
kernel notification.

On some embedded systems, there is no hotplug program. These 2
patches introduce a new EAL option "--create-uio-dev" that tells
the DPDK to create the /dev/uioX devices using mknod().

Olivier Matz (2):
  pci: split the function providing uio device and mappings
  pci: add option --create-uio-dev to run without hotplug

 lib/librte_eal/linuxapp/eal/eal.c                  |   6 +
 lib/librte_eal/linuxapp/eal/eal_pci.c              | 127 +++++++++++++++------
 .../linuxapp/eal/include/eal_internal_cfg.h        |   1 +
 3 files changed, 102 insertions(+), 32 deletions(-)

-- 
1.8.4.rc3

[PATCH 1/2] pci: split the function providing uio device and mappings

[Olivier Matz]

Add a new function pci_get_uio_dev() that parses /sys/bus/pci/devices
to get the uio device associated with a PCI device. This patch just
moves some code that was in pci_uio_map_resource() in the new function
without any functional change.

Thanks to this change, the next commit will be easier to understand.
Moreover it improves readability: having smaller functions help to
understand what pci_uio_map_resource() does.

Signed-off-by: Olivier Matz <olivier.matz-pdR9zngts4EAvxtiuMwx3w@public.gmane.org>
---
 lib/librte_eal/linuxapp/eal/eal_pci.c | 82 +++++++++++++++++++++--------------
 1 file changed, 50 insertions(+), 32 deletions(-)

diff --git a/lib/librte_eal/linuxapp/eal/eal_pci.c b/lib/librte_eal/linuxapp/eal/eal_pci.c
index 37ee6f1..1039777 100644
--- a/lib/librte_eal/linuxapp/eal/eal_pci.c
+++ b/lib/librte_eal/linuxapp/eal/eal_pci.c
@@ -460,34 +460,20 @@ pci_uio_map_secondary(struct rte_pci_device *dev)
 	return -1;
 }
 
-/* map the PCI resource of a PCI device in virtual memory */
-static int
-pci_uio_map_resource(struct rte_pci_device *dev)
+/*
+ * Return the uioX char device used for a pci device. On success, return
+ * the UIO number and fill dstbuf string with the path of the device in
+ * sysfs. On error, return a negative value. In this case dstbuf is
+ * invalid.
+ */
+static int pci_get_uio_dev(struct rte_pci_device *dev, char *dstbuf,
+			   unsigned int buflen)
 {
-	int i, j;
+	struct rte_pci_addr *loc = &dev->addr;
+	unsigned int uio_num;
 	struct dirent *e;
 	DIR *dir;
 	char dirname[PATH_MAX];
-	char filename[PATH_MAX];
-	char dirname2[PATH_MAX];
-	char devname[PATH_MAX]; /* contains the /dev/uioX */
-	void *mapaddr;
-	unsigned uio_num;
-	unsigned long start,size;
-	uint64_t phaddr;
-	uint64_t offset;
-	uint64_t pagesz;
-	ssize_t nb_maps;
-	struct rte_pci_addr *loc = &dev->addr;
-	struct uio_resource *uio_res;
-	struct uio_map *maps;
-
-	dev->intr_handle.fd = -1;
-
-	/* secondary processes - use already recorded details */
-	if ((rte_eal_process_type() != RTE_PROC_PRIMARY) &&
-		(dev->id.vendor_id != PCI_VENDOR_ID_QUMRANET))
-		return (pci_uio_map_secondary(dev));
 
 	/* depending on kernel version, uio can be located in uio/uioX
 	 * or uio:uioX */
@@ -525,8 +511,7 @@ pci_uio_map_resource(struct rte_pci_device *dev)
 		errno = 0;
 		uio_num = strtoull(e->d_name + shortprefix_len, &endptr, 10);
 		if (errno == 0 && endptr != (e->d_name + shortprefix_len)) {
-			rte_snprintf(dirname2, sizeof(dirname2),
-				 "%s/uio%u", dirname, uio_num);
+			rte_snprintf(dstbuf, buflen, "%s/uio%u", dirname, uio_num);
 			break;
 		}
 
@@ -534,15 +519,48 @@ pci_uio_map_resource(struct rte_pci_device *dev)
 		errno = 0;
 		uio_num = strtoull(e->d_name + longprefix_len, &endptr, 10);
 		if (errno == 0 && endptr != (e->d_name + longprefix_len)) {
-			rte_snprintf(dirname2, sizeof(dirname2),
-				 "%s/uio:uio%u", dirname, uio_num);
+			rte_snprintf(dstbuf, buflen, "%s/uio:uio%u", dirname, uio_num);
 			break;
 		}
 	}
 	closedir(dir);
 
 	/* No uio resource found */
-	if (e == NULL) {
+	if (e == NULL)
+		return -1;
+
+	return 0;
+}
+
+/* map the PCI resource of a PCI device in virtual memory */
+static int
+pci_uio_map_resource(struct rte_pci_device *dev)
+{
+	int i, j;
+	char dirname[PATH_MAX];
+	char filename[PATH_MAX];
+	char devname[PATH_MAX]; /* contains the /dev/uioX */
+	void *mapaddr;
+	int uio_num;
+	unsigned long start,size;
+	uint64_t phaddr;
+	uint64_t offset;
+	uint64_t pagesz;
+	ssize_t nb_maps;
+	struct rte_pci_addr *loc = &dev->addr;
+	struct uio_resource *uio_res;
+	struct uio_map *maps;
+
+	dev->intr_handle.fd = -1;
+
+	/* secondary processes - use already recorded details */
+	if ((rte_eal_process_type() != RTE_PROC_PRIMARY) &&
+	    (dev->id.vendor_id != PCI_VENDOR_ID_QUMRANET))
+		return (pci_uio_map_secondary(dev));
+
+	/* find uio resource */
+	uio_num = pci_get_uio_dev(dev, dirname, sizeof(dirname));
+	if (uio_num < 0) {
 		RTE_LOG(WARNING, EAL, "  "PCI_PRI_FMT" not managed by UIO driver, "
 				"skipping\n", loc->domain, loc->bus, loc->devid, loc->function);
 		return -1;
@@ -551,7 +569,7 @@ pci_uio_map_resource(struct rte_pci_device *dev)
 	if(dev->id.vendor_id == PCI_VENDOR_ID_QUMRANET) {
 		/* get portio size */
 		rte_snprintf(filename, sizeof(filename),
-			 "%s/portio/port0/size", dirname2);
+			 "%s/portio/port0/size", dirname);
 		if (eal_parse_sysfs_value(filename, &size) < 0) {
 			RTE_LOG(ERR, EAL, "%s(): cannot parse size\n",
 				__func__);
@@ -560,7 +578,7 @@ pci_uio_map_resource(struct rte_pci_device *dev)
 
 		/* get portio start */
 		rte_snprintf(filename, sizeof(filename),
-			 "%s/portio/port0/start", dirname2);
+			 "%s/portio/port0/start", dirname);
 		if (eal_parse_sysfs_value(filename, &start) < 0) {
 			RTE_LOG(ERR, EAL, "%s(): cannot parse portio start\n",
 				__func__);
@@ -585,7 +603,7 @@ pci_uio_map_resource(struct rte_pci_device *dev)
 	memcpy(&uio_res->pci_addr, &dev->addr, sizeof(uio_res->pci_addr));
 
 	/* collect info about device mappings */
-	if ((nb_maps = pci_uio_get_mappings(dirname2, uio_res->maps,
+	if ((nb_maps = pci_uio_get_mappings(dirname, uio_res->maps,
 			sizeof (uio_res->maps) / sizeof (uio_res->maps[0])))
 			< 0)
 		return (nb_maps);
-- 
1.8.4.rc3

Re: [PATCH 1/2] pci: split the function providing uio device and mappings

[Thomas Monjalon]

24/01/2014 16:31, Olivier Matz:
> Add a new function pci_get_uio_dev() that parses /sys/bus/pci/devices
> to get the uio device associated with a PCI device. This patch just
> moves some code that was in pci_uio_map_resource() in the new function
> without any functional change.
> 
> Thanks to this change, the next commit will be easier to understand.
> Moreover it improves readability: having smaller functions help to
> understand what pci_uio_map_resource() does.
> 
> Signed-off-by: Olivier Matz <olivier.matz-pdR9zngts4EAvxtiuMwx3w@public.gmane.org>

Acked and applied.

-- 
Thomas

[PATCH 2/2] pci: add option --create-uio-dev to run without hotplug

[Olivier Matz]

When the user specifies --create-uio-dev in dpdk eal start options, the
DPDK will create the /dev/uioX instead of waiting that a program does it
(which is usually hotplug).

This option is useful in embedded environments where there is no hotplug
to do the work.

Signed-off-by: Olivier Matz <olivier.matz-pdR9zngts4EAvxtiuMwx3w@public.gmane.org>
---
 lib/librte_eal/linuxapp/eal/eal.c                  |  6 +++
 lib/librte_eal/linuxapp/eal/eal_pci.c              | 47 +++++++++++++++++++++-
 .../linuxapp/eal/include/eal_internal_cfg.h        |  1 +
 3 files changed, 53 insertions(+), 1 deletion(-)

diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
index bd20331..9168b3f 100644
--- a/lib/librte_eal/linuxapp/eal/eal.c
+++ b/lib/librte_eal/linuxapp/eal/eal.c
@@ -94,6 +94,7 @@
 #define OPT_SOCKET_MEM  "socket-mem"
 #define OPT_USE_DEVICE  "use-device"
 #define OPT_SYSLOG      "syslog"
+#define OPT_CREATE_UIO_DEV "create-uio-dev"
 
 #define RTE_EAL_BLACKLIST_SIZE	0x100
 
@@ -357,6 +358,7 @@ eal_usage(const char *prgname)
 	       "  --"OPT_NO_PCI"   : disable pci\n"
 	       "  --"OPT_NO_HPET"  : disable hpet\n"
 	       "  --"OPT_NO_SHCONF": no shared config (mmap'd files)\n"
+	       "  --"OPT_CREATE_UIO_DEV": create /dev/uioX (usually done by hotplug)\n"
 	       "\n",
 	       prgname);
 	/* Allow the application to print its usage message too if hook is set */
@@ -608,6 +610,7 @@ eal_parse_args(int argc, char **argv)
 		{OPT_SOCKET_MEM, 1, 0, 0},
 		{OPT_USE_DEVICE, 1, 0, 0},
 		{OPT_SYSLOG, 1, NULL, 0},
+		{OPT_CREATE_UIO_DEV, 1, NULL, 0},
 		{0, 0, 0, 0}
 	};
 	struct shared_driver *solib;
@@ -747,6 +750,9 @@ eal_parse_args(int argc, char **argv)
 					return -1;
 				}
 			}
+			else if (!strcmp(lgopts[option_index].name, OPT_CREATE_UIO_DEV)) {
+				internal_config.create_uio_dev = 1;
+			}
 			break;
 
 		default:
diff --git a/lib/librte_eal/linuxapp/eal/eal_pci.c b/lib/librte_eal/linuxapp/eal/eal_pci.c
index 1039777..af5a8b9 100644
--- a/lib/librte_eal/linuxapp/eal/eal_pci.c
+++ b/lib/librte_eal/linuxapp/eal/eal_pci.c
@@ -460,6 +460,47 @@ pci_uio_map_secondary(struct rte_pci_device *dev)
 	return -1;
 }
 
+static int pci_mknod_uio_dev(const char *sysfs_uio_path, unsigned uio_num)
+{
+	FILE *f;
+	char filename[PATH_MAX];
+	int ret;
+	unsigned major, minor;
+	dev_t dev;
+
+	/* get the name of the sysfs file that contains the major and minor
+	 * of the uio device and read its content */
+	rte_snprintf(filename, sizeof(filename), "%s/dev", sysfs_uio_path);
+
+	f = fopen(filename, "r");
+	if (f == NULL) {
+		RTE_LOG(ERR, EAL, "%s(): cannot open sysfs to get major:minor\n",
+			__func__);
+		return -1;
+	}
+
+	ret = fscanf(f, "%d:%d", &major, &minor);
+	if (ret != 2) {
+		RTE_LOG(ERR, EAL, "%s(): cannot parse sysfs to get major:minor\n",
+			__func__);
+		fclose(f);
+		return -1;
+	}
+	fclose(f);
+
+	/* create the char device "mknod /dev/uioX c major minor" */
+	rte_snprintf(filename, sizeof(filename), "/dev/uio%u", uio_num);
+	dev = makedev(major, minor);
+	ret = mknod(filename, S_IFCHR | S_IRUSR | S_IWUSR, dev);
+	if (f == NULL) {
+		RTE_LOG(ERR, EAL, "%s(): mknod() failed %s\n",
+			__func__, strerror(errno));
+		return -1;
+	}
+
+	return ret;
+}
+
 /*
  * Return the uioX char device used for a pci device. On success, return
  * the UIO number and fill dstbuf string with the path of the device in
@@ -529,7 +570,11 @@ static int pci_get_uio_dev(struct rte_pci_device *dev, char *dstbuf,
 	if (e == NULL)
 		return -1;
 
-	return 0;
+	/* create uio device if we've been asked to */
+	if (internal_config.create_uio_dev && pci_mknod_uio_dev(dstbuf, uio_num) < 0)
+		RTE_LOG(WARNING, EAL, "Cannot create /dev/uio%u\n", uio_num);
+
+	return uio_num;
 }
 
 /* map the PCI resource of a PCI device in virtual memory */
diff --git a/lib/librte_eal/linuxapp/eal/include/eal_internal_cfg.h b/lib/librte_eal/linuxapp/eal/include/eal_internal_cfg.h
index 45ec058..649bd2b 100644
--- a/lib/librte_eal/linuxapp/eal/include/eal_internal_cfg.h
+++ b/lib/librte_eal/linuxapp/eal/include/eal_internal_cfg.h
@@ -68,6 +68,7 @@ struct internal_config {
 	volatile unsigned vmware_tsc_map; /**< true to use VMware TSC mapping
 										* instead of native TSC */
 	volatile unsigned no_shconf;      /**< true if there is no shared config */
+	volatile unsigned create_uio_dev; /**< true to create /dev/uioX devices */
 	volatile enum rte_proc_type_t process_type; /**< multi-process proc type */
 	/** true to try allocating memory on specific sockets */
 	volatile unsigned force_sockets;
-- 
1.8.4.rc3

Re: [PATCH 2/2] pci: add option --create-uio-dev to run without hotplug

[Thomas Monjalon]

24/01/2014 16:31, Olivier Matz:
> --- a/lib/librte_eal/linuxapp/eal/eal.c
> +++ b/lib/librte_eal/linuxapp/eal/eal.c
> @@ -357,6 +358,7 @@ eal_usage(const char *prgname)
>  	       "  --"OPT_NO_PCI"   : disable pci\n"
>  	       "  --"OPT_NO_HPET"  : disable hpet\n"
>  	       "  --"OPT_NO_SHCONF": no shared config (mmap'd files)\n"
> +	       "  --"OPT_CREATE_UIO_DEV": create /dev/uioX (usually done by
> hotplug)\n" "\n",

The option is added in section "EAL options for DEBUG use only".
Shouldn't be in the general section "EAL options" ?

-- 
Thomas

[PATCH 2/2 v2] pci: add option --create-uio-dev to run without hotplug

[Olivier Matz]

When the user specifies --create-uio-dev in dpdk eal start options, the
DPDK will create the /dev/uioX instead of waiting that a program does it
(which is usually hotplug).

This option is useful in embedded environments where there is no hotplug
to do the work.

Signed-off-by: Olivier Matz <olivier.matz-pdR9zngts4EAvxtiuMwx3w@public.gmane.org>
---
 lib/librte_eal/linuxapp/eal/eal.c                  |  6 +++
 lib/librte_eal/linuxapp/eal/eal_pci.c              | 47 +++++++++++++++++++++-
 .../linuxapp/eal/include/eal_internal_cfg.h        |  1 +
 3 files changed, 53 insertions(+), 1 deletion(-)

Hi Thomas,

Thanks for your comment, I moved the option help above.

diff --git a/lib/librte_eal/linuxapp/eal/eal.c b/lib/librte_eal/linuxapp/eal/eal.c
index bd20331..e402fee 100644
--- a/lib/librte_eal/linuxapp/eal/eal.c
+++ b/lib/librte_eal/linuxapp/eal/eal.c
@@ -94,6 +94,7 @@
 #define OPT_SOCKET_MEM  "socket-mem"
 #define OPT_USE_DEVICE  "use-device"
 #define OPT_SYSLOG      "syslog"
+#define OPT_CREATE_UIO_DEV "create-uio-dev"
 
 #define RTE_EAL_BLACKLIST_SIZE	0x100
 
@@ -352,6 +353,7 @@ eal_usage(const char *prgname)
 	       "               [NOTE: Cannot be used with -b option]\n"
 	       "  --"OPT_VMWARE_TSC_MAP": use VMware TSC map instead of "
 	    		   "native RDTSC\n"
+	       "  --"OPT_CREATE_UIO_DEV": create /dev/uioX (usually done by hotplug)\n"
 	       "\nEAL options for DEBUG use only:\n"
 	       "  --"OPT_NO_HUGE"  : use malloc instead of hugetlbfs\n"
 	       "  --"OPT_NO_PCI"   : disable pci\n"
@@ -608,6 +610,7 @@ eal_parse_args(int argc, char **argv)
 		{OPT_SOCKET_MEM, 1, 0, 0},
 		{OPT_USE_DEVICE, 1, 0, 0},
 		{OPT_SYSLOG, 1, NULL, 0},
+		{OPT_CREATE_UIO_DEV, 1, NULL, 0},
 		{0, 0, 0, 0}
 	};
 	struct shared_driver *solib;
@@ -747,6 +750,9 @@ eal_parse_args(int argc, char **argv)
 					return -1;
 				}
 			}
+			else if (!strcmp(lgopts[option_index].name, OPT_CREATE_UIO_DEV)) {
+				internal_config.create_uio_dev = 1;
+			}
 			break;
 
 		default:
diff --git a/lib/librte_eal/linuxapp/eal/eal_pci.c b/lib/librte_eal/linuxapp/eal/eal_pci.c
index 1039777..af5a8b9 100644
--- a/lib/librte_eal/linuxapp/eal/eal_pci.c
+++ b/lib/librte_eal/linuxapp/eal/eal_pci.c
@@ -460,6 +460,47 @@ pci_uio_map_secondary(struct rte_pci_device *dev)
 	return -1;
 }
 
+static int pci_mknod_uio_dev(const char *sysfs_uio_path, unsigned uio_num)
+{
+	FILE *f;
+	char filename[PATH_MAX];
+	int ret;
+	unsigned major, minor;
+	dev_t dev;
+
+	/* get the name of the sysfs file that contains the major and minor
+	 * of the uio device and read its content */
+	rte_snprintf(filename, sizeof(filename), "%s/dev", sysfs_uio_path);
+
+	f = fopen(filename, "r");
+	if (f == NULL) {
+		RTE_LOG(ERR, EAL, "%s(): cannot open sysfs to get major:minor\n",
+			__func__);
+		return -1;
+	}
+
+	ret = fscanf(f, "%d:%d", &major, &minor);
+	if (ret != 2) {
+		RTE_LOG(ERR, EAL, "%s(): cannot parse sysfs to get major:minor\n",
+			__func__);
+		fclose(f);
+		return -1;
+	}
+	fclose(f);
+
+	/* create the char device "mknod /dev/uioX c major minor" */
+	rte_snprintf(filename, sizeof(filename), "/dev/uio%u", uio_num);
+	dev = makedev(major, minor);
+	ret = mknod(filename, S_IFCHR | S_IRUSR | S_IWUSR, dev);
+	if (f == NULL) {
+		RTE_LOG(ERR, EAL, "%s(): mknod() failed %s\n",
+			__func__, strerror(errno));
+		return -1;
+	}
+
+	return ret;
+}
+
 /*
  * Return the uioX char device used for a pci device. On success, return
  * the UIO number and fill dstbuf string with the path of the device in
@@ -529,7 +570,11 @@ static int pci_get_uio_dev(struct rte_pci_device *dev, char *dstbuf,
 	if (e == NULL)
 		return -1;
 
-	return 0;
+	/* create uio device if we've been asked to */
+	if (internal_config.create_uio_dev && pci_mknod_uio_dev(dstbuf, uio_num) < 0)
+		RTE_LOG(WARNING, EAL, "Cannot create /dev/uio%u\n", uio_num);
+
+	return uio_num;
 }
 
 /* map the PCI resource of a PCI device in virtual memory */
diff --git a/lib/librte_eal/linuxapp/eal/include/eal_internal_cfg.h b/lib/librte_eal/linuxapp/eal/include/eal_internal_cfg.h
index 45ec058..649bd2b 100644
--- a/lib/librte_eal/linuxapp/eal/include/eal_internal_cfg.h
+++ b/lib/librte_eal/linuxapp/eal/include/eal_internal_cfg.h
@@ -68,6 +68,7 @@ struct internal_config {
 	volatile unsigned vmware_tsc_map; /**< true to use VMware TSC mapping
 										* instead of native TSC */
 	volatile unsigned no_shconf;      /**< true if there is no shared config */
+	volatile unsigned create_uio_dev; /**< true to create /dev/uioX devices */
 	volatile enum rte_proc_type_t process_type; /**< multi-process proc type */
 	/** true to try allocating memory on specific sockets */
 	volatile unsigned force_sockets;
-- 
1.8.4.rc3

Re: [PATCH 2/2 v2] pci: add option --create-uio-dev to run without hotplug

[Thomas Monjalon]

28/01/2014 16:26, Olivier Matz:
> When the user specifies --create-uio-dev in dpdk eal start options, the
> DPDK will create the /dev/uioX instead of waiting that a program does it
> (which is usually hotplug).
> 
> This option is useful in embedded environments where there is no hotplug
> to do the work.
> 
> Signed-off-by: Olivier Matz <olivier.matz-pdR9zngts4EAvxtiuMwx3w@public.gmane.org>

Acked and applied.
Thank you
-- 
Thomas