Skip to contentSkip to conversation

Add character device interface to remoteproc

PATCH Superseded Rishabh Bhatnagar 5 participants 13 messages Mar 26, 2020 – Mar 31, 2020 Remote processor (remoteproc)

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

cover Add character device interface to remoteproc

This patch series adds a character device interface to remoteproc framework. This interface can be used by userspace clients in order to boot and shutdown the remote processor.
Currently there is only a sysfs interface which the userspace clients can use. If a usersapce application crashes after booting the remote processor through the sysfs interface the remote processor does not get any indication about the crash. It might still assume that the application is running.
For example modem uses remotefs service to data from disk/flash memory. If the remotefs service crashes, modem still keeps on requesting data which might lead to crash on modem. Even if the service is restarted the file handles modem requested previously would become stale. Adding a character device interface makes the remote processor tightly coupled with the user space application. A crash of the application leads to a close on the file descriptors therefore shutting down the remoteproc.

Rishabh Bhatnagar (2):
remoteproc: Add userspace char device driver
remoteproc: core: Register the character device interface

 drivers/remoteproc/Makefile               |   1 +
 drivers/remoteproc/remoteproc_core.c      |   8 +-
 drivers/remoteproc/remoteproc_internal.h  |   3 +
 drivers/remoteproc/remoteproc_userspace.c | 126 ++++++++++++++++++++++++++++++
 include/linux/remoteproc.h                |   2 +
 5 files changed, 139 insertions(+), 1 deletion(-)
 create mode 100644 drivers/remoteproc/remoteproc_userspace.c
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#11 lore raw

PATCH 2/2 remoteproc: core: Register the character device interface

Add the character device during rproc_add. This would create a character device node at /dev/subsys_<rproc_name>. Userspace applications can interact with the remote processor using this interface rather than using sysfs node. To distinguish between different remote processor nodes the device name has been changed to include the rproc name appended to "rproc_" string.

Change-Id I2114f77f8d2b5fd97e281021ec9905ef5c2fb54c
---
 drivers/remoteproc/remoteproc_core.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
drivers/remoteproc/remoteproc_core.c +9 -1
index 097f33e..f657983 100644
@@ -1907,6 +1907,12 @@ int rproc_add(struct rproc *rproc)
 	struct device *dev = &rproc->dev;
 	int ret;
 
+	ret = rproc_char_device_add(rproc);
+	if (ret) {
+		pr_err("error while adding character device\n");
+		return ret;
+	}
+
 	ret = device_add(dev);
 	if (ret < 0)
 		return ret;
@@ -2044,7 +2050,7 @@ struct rproc *rproc_alloc(struct device *dev, const char *name,
 		return NULL;
 	}
 
-	dev_set_name(&rproc->dev, "remoteproc%d", rproc->index);
+	dev_set_name(&rproc->dev, "rproc_%s", rproc->name);
 
 	atomic_set(&rproc->power, 0);
 
@@ -2220,6 +2226,7 @@ static int __init remoteproc_init(void)
 {
 	rproc_init_sysfs();
 	rproc_init_debugfs();
+	rproc_init_cdev();
 
 	return 0;
 }
@@ -2231,6 +2238,7 @@ static void __exit remoteproc_exit(void)
 
 	rproc_exit_debugfs();
 	rproc_exit_sysfs();
+	rproc_exit_cdev();
 }
 module_exit(remoteproc_exit);
 
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#1 lore raw

PATCH 1/2 remoteproc: Add userspace char device driver

In reply to Rishabh Bhatnagar

Add the driver for creating the character device interface for userspace applications. The character device interface can be used in order to boot up and shutdown the remote processor. This might be helpful for remote processors that are booted by userspace applications and need to shutdown when the application crahes/shutsdown.

Change-Id If23c8986272bb7c943eb76665f127ff706f12394
---
 drivers/remoteproc/Makefile               |  1 +
 drivers/remoteproc/remoteproc_internal.h  |  6 +++
 drivers/remoteproc/remoteproc_userspace.c | 90 +++++++++++++++++++++++++++++++
 include/linux/remoteproc.h                |  2 +
 4 files changed, 99 insertions(+)
 create mode 100644 drivers/remoteproc/remoteproc_userspace.c
drivers/remoteproc/Makefile +1 -0
index e30a1b1..facb3fa 100644
@@ -7,6 +7,7 @@ obj-$(CONFIG_REMOTEPROC)		+= remoteproc.o
 remoteproc-y				:= remoteproc_core.o
 remoteproc-y				+= remoteproc_debugfs.o
 remoteproc-y				+= remoteproc_sysfs.o
+remoteproc-y				+= remoteproc_userspace.o
 remoteproc-y				+= remoteproc_virtio.o
 remoteproc-y				+= remoteproc_elf_loader.o
 obj-$(CONFIG_IMX_REMOTEPROC)		+= imx_rproc.o
drivers/remoteproc/remoteproc_internal.h +6 -0
index 493ef92..97513ba 100644
@@ -47,6 +47,9 @@ struct dentry *rproc_create_trace_file(const char *name, struct rproc *rproc,
 int rproc_init_sysfs(void);
 void rproc_exit_sysfs(void);
 
+void rproc_init_cdev(void);
+void rproc_exit_cdev(void);
+
 void rproc_free_vring(struct rproc_vring *rvring);
 int rproc_alloc_vring(struct rproc_vdev *rvdev, int i);
 
@@ -63,6 +66,9 @@ struct resource_table *rproc_elf_find_loaded_rsc_table(struct rproc *rproc,
 struct rproc_mem_entry *
 rproc_find_carveout_by_name(struct rproc *rproc, const char *name, ...);
 
+/* from remoteproc_userspace.c */
+int rproc_char_device_add(struct rproc *rproc);
+
 static inline
 int rproc_fw_sanity_check(struct rproc *rproc, const struct firmware *fw)
 {
drivers/remoteproc/remoteproc_userspace.c new file +90 -0

90 lines not on this page · raw message

include/linux/remoteproc.h +2 -0
index 16ad666..c4ca796 100644
@@ -37,6 +37,7 @@
 
 #include <linux/types.h>
 #include <linux/mutex.h>
+#include <linux/cdev.h>
 #include <linux/virtio.h>
 #include <linux/completion.h>
 #include <linux/idr.h>
@@ -514,6 +515,7 @@ struct rproc {
 	bool auto_boot;
 	struct list_head dump_segments;
 	int nb_vdev;
+	struct cdev char_dev;
 };
 
 /**
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#2 lore raw

Re: [PATCH 1/2] remoteproc: Add userspace char device driver

Hi Rishabh,

While being interesting to have a such a userspace interface, I have some remarks.

Quoted 82 lines · ----- On 26 Mar, 2020, at 17:50, Rishabh Bhatnagar [email protected]…

Add the driver for creating the character device interface for userspace applications. The character device interface can be used in order to boot up and shutdown the remote processor. This might be helpful for remote processors that are booted by userspace applications and need to shutdown when the application crahes/shutsdown.

Change-Id If23c8986272bb7c943eb76665f127ff706f12394

---

drivers/remoteproc/Makefile               |  1 +
drivers/remoteproc/remoteproc_internal.h  |  6 +++
drivers/remoteproc/remoteproc_userspace.c | 90 +++++++++++++++++++++++++++++++
include/linux/remoteproc.h                |  2 +
4 files changed, 99 insertions(+)

create mode 100644 drivers/remoteproc/remoteproc_userspace.c

drivers/remoteproc/Makefile +0 -0
index e30a1b1..facb3fa 100644
@@ -7,6 +7,7 @@ obj-$(CONFIG_REMOTEPROC)		+= remoteproc.o

remoteproc-y := remoteproc_core.o
remoteproc-y += remoteproc_debugfs.o
remoteproc-y += remoteproc_sysfs.o
+remoteproc-y += remoteproc_userspace.o
remoteproc-y += remoteproc_virtio.o
remoteproc-y += remoteproc_elf_loader.o
obj-$(CONFIG_IMX_REMOTEPROC) += imx_rproc.o

drivers/remoteproc/remoteproc_internal.h +3 -0
index 493ef92..97513ba 100644
@@ -47,6 +47,9 @@ struct dentry *rproc_create_trace_file(const char *name,
 struct rproc *rproc,
 int rproc_init_sysfs(void);
 void rproc_exit_sysfs(void);
 
+void rproc_init_cdev(void);
+void rproc_exit_cdev(void);
+
 void rproc_free_vring(struct rproc_vring *rvring);
 int rproc_alloc_vring(struct rproc_vdev *rvdev, int i);
@@ -63,6 +66,9 @@ struct resource_table *rproc_elf_find_loaded_rsc_table(struct
 rproc *rproc,

struct rproc_mem_entry *
rproc_find_carveout_by_name(struct rproc *rproc, const char *name, ...);

drivers/remoteproc/remoteproc_internal.h +3 -0
+/* from remoteproc_userspace.c */
+int rproc_char_device_add(struct rproc *rproc);
+

static inline
int rproc_fw_sanity_check(struct rproc *rproc, const struct firmware *fw) {

drivers/remoteproc/remoteproc_userspace.c new file +19 -0
new file mode 100644
index 0000000..2ef7679
@@ -0,0 +1,90 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Character device interface driver for Remoteproc framework.
+ *
+ * Copyright (c) 2020, The Linux Foundation. All rights reserved.
+ */
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/cdev.h>
+#include <linux/mutex.h>
+#include <linux/remoteproc.h>
+
+#include "remoteproc_internal.h"
+
+#define NUM_RPROC_DEVICES	64
+static dev_t rproc_cdev;
+static DEFINE_IDA(cdev_minor_ida);
+

drivers/remoteproc/remoteproc_userspace.c line 29

Quoted 10 lines
drivers/remoteproc/remoteproc_userspace.c new file +10 -0
+static int rproc_open(struct inode *inode, struct file *file)
+{
+	struct rproc *rproc;
+
+	rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+	if (!rproc)
+		return -EINVAL;
+
+	return rproc_boot(rproc);
+}

What happens if multiple user open this chardev ? Apparently, rproc_boot returns 0 if already powered_up, so the next user won't know what is the state of the rproc.
Exclusive access could probably be a good idea.

Quoted 21 lines
drivers/remoteproc/remoteproc_userspace.c new file +21 -0
+
+static int rproc_release(struct inode *inode, struct file *file)
+{
+	struct rproc *rproc;
+
+	rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+	if (!rproc)
+		return -EINVAL;
+
+	rproc_shutdown(rproc);
+
+	return 0;
+}
+
+static const struct file_operations rproc_fops = {
+	.open = rproc_open,
+	.release = rproc_release,
+};
+
+int rproc_char_device_add(struct rproc *rproc)
+{

drivers/remoteproc/remoteproc_userspace.c line 60

Quoted 10 lines
drivers/remoteproc/remoteproc_userspace.c new file +10 -0
+	int ret, minor;
+	dev_t cdevt;
+
+	minor = ida_simple_get(&cdev_minor_ida, 0, NUM_RPROC_DEVICES,
+			GFP_KERNEL);
+	if (minor < 0) {
+	pr_err("%s: No more minor numbers left! rc:%d\n", __func__,
+							minor);
+		return -ENODEV;
+	}

How can you make the link between the chardev and the device instance ? In our case, we have several remoteproc instances and thus we will end up having multiple chardev.

Regards,

Clément

Quoted 53 lines
drivers/remoteproc/remoteproc_userspace.c new file +30 -0
+
+	cdev_init(&rproc->char_dev, &rproc_fops);
+	rproc->char_dev.owner = THIS_MODULE;
+
+	cdevt = MKDEV(MAJOR(rproc_cdev), minor);
+	ret = cdev_add(&rproc->char_dev, cdevt, 1);
+	if (ret < 0)
+		ida_simple_remove(&cdev_minor_ida, minor);
+
+	rproc->dev.devt = cdevt;
+
+	return ret;
+}
+
+void __init rproc_init_cdev(void)
+{
+	int ret;
+
+	ret = alloc_chrdev_region(&rproc_cdev, 0, NUM_RPROC_DEVICES, "rproc");
+	if (ret < 0) {
+		pr_err("Failed to alloc rproc_cdev region, err %d\n", ret);
+		return;
+	}
+}
+
+void __exit rproc_exit_cdev(void)
+{
+	unregister_chrdev_region(MKDEV(MAJOR(rproc_cdev), 0),
+				NUM_RPROC_DEVICES);
+}
include/linux/remoteproc.h +2 -1
index 16ad666..c4ca796 100644
@@ -37,6 +37,7 @@
 
 #include <linux/types.h>
 #include <linux/mutex.h>
+#include <linux/cdev.h>
 #include <linux/virtio.h>
 #include <linux/completion.h>
 #include <linux/idr.h>
@@ -514,6 +515,7 @@ struct rproc {
 	bool auto_boot;
 	struct list_head dump_segments;
 	int nb_vdev;
+	struct cdev char_dev;
 };
 
 /**
--

The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project

#5 lore raw

Re: [PATCH 1/2] remoteproc: Add userspace char device driver

In reply to Rishabh Bhatnagar
Thank you for the patch! Yet something to improve:

Hi Rishabh,

Thank you for the patch! Yet something to improve:

[auto build test ERROR on linus/master]
[also build test ERROR on linux/master v5.6-rc7]
[cannot apply to next-20200326]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system. BTW, we also suggest to use '--base' option to specify the base tree in git format-patch, please see https://stackoverflow.com/a/37406982]

url:    https://github.com/0day-ci/linux/commits/Rishabh-Bhatnagar/Add-character-device-interface-to-remoteproc/20200327-062958
base:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git 9420e8ade4353a6710908ffafa23ecaf1caa0123
config: m68k-allmodconfig (attached as .config)
compiler: m68k-linux-gcc (GCC) 9.2.0

reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
GCC_VERSION=9.2.0 make.cross ARCH=m68k

If you fix the issue, kindly add following tag

All errors (new ones prefixed by >>):

>> drivers/remoteproc/remoteproc_userspace.c:31:12: error: conflicting types for 'rproc_release'
      31 | static int rproc_release(struct inode *inode, struct file *file)
         |            ^~~~~~~~~~~~~
   In file included from drivers/remoteproc/remoteproc_userspace.c:14:
   drivers/remoteproc/remoteproc_internal.h:28:6: note: previous declaration of 'rproc_release' was here
      28 | void rproc_release(struct kref *kref);
         |      ^~~~~~~~~~~~~

vim +/rproc_release +31 drivers/remoteproc/remoteproc_userspace.c

    30	
  > 31	static int rproc_release(struct inode *inode, struct file *file)
    32	{
    33		struct rproc *rproc;
    34	
    35		rproc = container_of(inode->i_cdev, struct rproc, char_dev);
    36		if (!rproc)
    37			return -EINVAL;
    38	
    39		rproc_shutdown(rproc);
    40	
    41		return 0;
    42	}
    43	

---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org

#3 lore raw

Re: [PATCH 1/2] remoteproc: Add userspace char device driver

In reply to Clément Leger
Quoted 100 lines · Clément Leger

Hi Rishabh,

While being interesting to have a such a userspace interface, I have some remarks.

----- On 26 Mar, 2020, at 17:50, Rishabh Bhatnagar

Quoted 92 lines · [email protected] wrote:

Add the driver for creating the character device interface for userspace applications. The character device interface can be used in order to boot up and shutdown the remote processor. This might be helpful for remote processors that are booted by userspace applications and need to shutdown when the application crahes/shutsdown.

Change-Id If23c8986272bb7c943eb76665f127ff706f12394

---

drivers/remoteproc/Makefile               |  1 +
drivers/remoteproc/remoteproc_internal.h  |  6 +++
drivers/remoteproc/remoteproc_userspace.c | 90 
+++++++++++++++++++++++++++++++
include/linux/remoteproc.h                |  2 +
4 files changed, 99 insertions(+)

create mode 100644 drivers/remoteproc/remoteproc_userspace.c

drivers/remoteproc/Makefile +0 -0
index e30a1b1..facb3fa 100644
@@ -7,6 +7,7 @@ obj-$(CONFIG_REMOTEPROC)		+= remoteproc.o

remoteproc-y := remoteproc_core.o
remoteproc-y += remoteproc_debugfs.o
remoteproc-y += remoteproc_sysfs.o
+remoteproc-y += remoteproc_userspace.o
remoteproc-y += remoteproc_virtio.o
remoteproc-y += remoteproc_elf_loader.o
obj-$(CONFIG_IMX_REMOTEPROC) += imx_rproc.o

drivers/remoteproc/remoteproc_internal.h +3 -0
index 493ef92..97513ba 100644
@@ -47,6 +47,9 @@ struct dentry *rproc_create_trace_file(const char 
 *name,
 struct rproc *rproc,
 int rproc_init_sysfs(void);
 void rproc_exit_sysfs(void);
 
+void rproc_init_cdev(void);
+void rproc_exit_cdev(void);
+
 void rproc_free_vring(struct rproc_vring *rvring);

int rproc_alloc_vring(struct rproc_vdev *rvdev, int i);

drivers/remoteproc/remoteproc_internal.h +0 -0
@@ -63,6 +66,9 @@ struct resource_table 
 *rproc_elf_find_loaded_rsc_table(struct
 rproc *rproc,

struct rproc_mem_entry *
rproc_find_carveout_by_name(struct rproc *rproc, const char *name, ...);

drivers/remoteproc/remoteproc_internal.h +3 -0
+/* from remoteproc_userspace.c */
+int rproc_char_device_add(struct rproc *rproc);
+

static inline
int rproc_fw_sanity_check(struct rproc *rproc, const struct firmware *fw) {

drivers/remoteproc/remoteproc_userspace.c new file +26 -0
new file mode 100644
index 0000000..2ef7679
@@ -0,0 +1,90 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Character device interface driver for Remoteproc framework.
+ *
+ * Copyright (c) 2020, The Linux Foundation. All rights reserved.
+ */
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/cdev.h>
+#include <linux/mutex.h>
+#include <linux/remoteproc.h>
+
+#include "remoteproc_internal.h"
+
+#define NUM_RPROC_DEVICES	64
+static dev_t rproc_cdev;
+static DEFINE_IDA(cdev_minor_ida);
+
+static int rproc_open(struct inode *inode, struct file *file)
+{
+	struct rproc *rproc;
+
+	rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+	if (!rproc)
+		return -EINVAL;
Quoted 8 lines

drivers/remoteproc/remoteproc_userspace.c line 29

Quoted 3 lines
drivers/remoteproc/remoteproc_userspace.c new file +3 -0
+
+	return rproc_boot(rproc);
+}

What happens if multiple user open this chardev ? Apparently, rproc_boot returns 0 if already powered_up, so the next user won't know what is the state of the rproc.
Exclusive access could probably be a good idea.

Since it is synchronized inside rproc_boot multiple users simultaneously calling open shouldn't be a problem. If it is one after the other then second caller will get result as 0 and assume that rproc booted successfully. That is the expected flow right?

Quoted 26 lines
Quoted 26 lines
drivers/remoteproc/remoteproc_userspace.c new file +25 -0
+
+static int rproc_release(struct inode *inode, struct file *file)
+{
+	struct rproc *rproc;
+
+	rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+	if (!rproc)
+		return -EINVAL;
+
+	rproc_shutdown(rproc);
+
+	return 0;
+}
+
+static const struct file_operations rproc_fops = {
+	.open = rproc_open,
+	.release = rproc_release,
+};
+
+int rproc_char_device_add(struct rproc *rproc)
+{
+	int ret, minor;
+	dev_t cdevt;
+
+	minor = ida_simple_get(&cdev_minor_ida, 0, NUM_RPROC_DEVICES,
Quoted 8 lines

drivers/remoteproc/remoteproc_userspace.c line 60

Quoted 6 lines
drivers/remoteproc/remoteproc_userspace.c new file +6 -0
+			GFP_KERNEL);
+	if (minor < 0) {
+	pr_err("%s: No more minor numbers left! rc:%d\n", __func__,
+							minor);
+		return -ENODEV;
+	}

How can you make the link between the chardev and the device instance ?

I do this rproc->dev.devt = cdevt. Let me know of there is a better way to do this?

Quoted 7 lines

In our case, we have several remoteproc instances and thus we will end up having multiple chardev.

Regards,

Clément

rproc_char_device_add will be called for each remoteproc that is added. So we will have one char dev per remoteproc.

Quoted 54 lines
Quoted 54 lines
drivers/remoteproc/remoteproc_userspace.c new file +19 -0
+
+	cdev_init(&rproc->char_dev, &rproc_fops);
+	rproc->char_dev.owner = THIS_MODULE;
+
+	cdevt = MKDEV(MAJOR(rproc_cdev), minor);
+	ret = cdev_add(&rproc->char_dev, cdevt, 1);
+	if (ret < 0)
+		ida_simple_remove(&cdev_minor_ida, minor);
+
+	rproc->dev.devt = cdevt;
+
+	return ret;
+}
+
+void __init rproc_init_cdev(void)
+{
+	int ret;
+
+	ret = alloc_chrdev_region(&rproc_cdev, 0, NUM_RPROC_DEVICES, 

"rproc");

drivers/remoteproc/remoteproc_userspace.c new file +11 -0
+	if (ret < 0) {
+		pr_err("Failed to alloc rproc_cdev region, err %d\n", ret);
+		return;
+	}
+}
+
+void __exit rproc_exit_cdev(void)
+{
+	unregister_chrdev_region(MKDEV(MAJOR(rproc_cdev), 0),
+				NUM_RPROC_DEVICES);
+}
include/linux/remoteproc.h +2 -1
index 16ad666..c4ca796 100644
@@ -37,6 +37,7 @@
 
 #include <linux/types.h>
 #include <linux/mutex.h>
+#include <linux/cdev.h>
 #include <linux/virtio.h>
 #include <linux/completion.h>
 #include <linux/idr.h>
@@ -514,6 +515,7 @@ struct rproc {
 	bool auto_boot;
 	struct list_head dump_segments;
 	int nb_vdev;
+	struct cdev char_dev;
 };
 
 /**
--

The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum, a Linux Foundation Collaborative Project

#4 lore raw

Re: [PATCH 1/2] remoteproc: Add userspace char device driver

Hi Rishabh,

Quoted 111 lines · ----- On 28 Mar, 2020, at 01:09, Rishabh Bhatnagar [email protected]…
Quoted 102 lines · Clément Leger

Hi Rishabh,

While being interesting to have a such a userspace interface, I have some remarks.

----- On 26 Mar, 2020, at 17:50, Rishabh Bhatnagar

Quoted 94 lines · [email protected] wrote:

Add the driver for creating the character device interface for userspace applications. The character device interface can be used in order to boot up and shutdown the remote processor. This might be helpful for remote processors that are booted by userspace applications and need to shutdown when the application crahes/shutsdown.

Change-Id If23c8986272bb7c943eb76665f127ff706f12394

---

drivers/remoteproc/Makefile               |  1 +
drivers/remoteproc/remoteproc_internal.h  |  6 +++
drivers/remoteproc/remoteproc_userspace.c | 90
+++++++++++++++++++++++++++++++
include/linux/remoteproc.h                |  2 +
4 files changed, 99 insertions(+)

create mode 100644 drivers/remoteproc/remoteproc_userspace.c

drivers/remoteproc/Makefile +0 -0
index e30a1b1..facb3fa 100644
@@ -7,6 +7,7 @@ obj-$(CONFIG_REMOTEPROC)		+= remoteproc.o

remoteproc-y := remoteproc_core.o
remoteproc-y += remoteproc_debugfs.o
remoteproc-y += remoteproc_sysfs.o
+remoteproc-y += remoteproc_userspace.o
remoteproc-y += remoteproc_virtio.o
remoteproc-y += remoteproc_elf_loader.o
obj-$(CONFIG_IMX_REMOTEPROC) += imx_rproc.o

drivers/remoteproc/remoteproc_internal.h +3 -0
index 493ef92..97513ba 100644
@@ -47,6 +47,9 @@ struct dentry *rproc_create_trace_file(const char
 *name,
 struct rproc *rproc,
 int rproc_init_sysfs(void);
 void rproc_exit_sysfs(void);
 
+void rproc_init_cdev(void);
+void rproc_exit_cdev(void);
+
 void rproc_free_vring(struct rproc_vring *rvring);

int rproc_alloc_vring(struct rproc_vdev *rvdev, int i);

drivers/remoteproc/remoteproc_internal.h +0 -0
@@ -63,6 +66,9 @@ struct resource_table
 *rproc_elf_find_loaded_rsc_table(struct
 rproc *rproc,

struct rproc_mem_entry * rproc_find_carveout_by_name(struct rproc *rproc, const char *name, ...);

drivers/remoteproc/remoteproc_internal.h +3 -0
+/* from remoteproc_userspace.c */
+int rproc_char_device_add(struct rproc *rproc);
+

static inline
int rproc_fw_sanity_check(struct rproc *rproc, const struct firmware *fw)
{

drivers/remoteproc/remoteproc_userspace.c new file +26 -0
new file mode 100644
index 0000000..2ef7679
@@ -0,0 +1,90 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Character device interface driver for Remoteproc framework.
+ *
+ * Copyright (c) 2020, The Linux Foundation. All rights reserved.
+ */
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/cdev.h>
+#include <linux/mutex.h>
+#include <linux/remoteproc.h>
+
+#include "remoteproc_internal.h"
+
+#define NUM_RPROC_DEVICES	64
+static dev_t rproc_cdev;
+static DEFINE_IDA(cdev_minor_ida);
+
+static int rproc_open(struct inode *inode, struct file *file)
+{
+	struct rproc *rproc;
+
+	rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+	if (!rproc)
+		return -EINVAL;
Quoted 8 lines

drivers/remoteproc/remoteproc_userspace.c line 29

Quoted 3 lines
drivers/remoteproc/remoteproc_userspace.c new file +3 -0
+
+	return rproc_boot(rproc);
+}

What happens if multiple user open this chardev ? Apparently, rproc_boot returns 0 if already powered_up, so the next user won't know what is the state of the rproc.
Exclusive access could probably be a good idea.

Quoted 4 lines

Since it is synchronized inside rproc_boot multiple users simultaneously calling open shouldn't be a problem. If it is one after the other then second caller will get result as 0 and assume that rproc booted successfully.

It will be the same for close, it will assume the rproc has been stopped ? But in fact it will still be running until the refcount is 0.

Quoted 1 line

That is the expected flow right?

I would expect only one caller to be successful, others should probably receive a EBUSY errno IMHO.

Quoted 34 lines
Quoted 26 lines
Quoted 26 lines
drivers/remoteproc/remoteproc_userspace.c new file +25 -0
+
+static int rproc_release(struct inode *inode, struct file *file)
+{
+	struct rproc *rproc;
+
+	rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+	if (!rproc)
+		return -EINVAL;
+
+	rproc_shutdown(rproc);
+
+	return 0;
+}
+
+static const struct file_operations rproc_fops = {
+	.open = rproc_open,
+	.release = rproc_release,
+};
+
+int rproc_char_device_add(struct rproc *rproc)
+{
+	int ret, minor;
+	dev_t cdevt;
+
+	minor = ida_simple_get(&cdev_minor_ida, 0, NUM_RPROC_DEVICES,
Quoted 8 lines

drivers/remoteproc/remoteproc_userspace.c line 60

Quoted 6 lines
drivers/remoteproc/remoteproc_userspace.c new file +6 -0
+			GFP_KERNEL);
+	if (minor < 0) {
+	pr_err("%s: No more minor numbers left! rc:%d\n", __func__,
+							minor);
+		return -ENODEV;
+	}

How can you make the link between the chardev and the device instance ?

Quoted 2 lines

I do this rproc->dev.devt = cdevt. Let me know of there is a better way to do this?

If this is sufficient to create a link in the sysfs, then it's ok but I'm no expert here.

Clément

Quoted 64 lines
Quoted 7 lines

In our case, we have several remoteproc instances and thus we will end up having multiple chardev.

Regards,

Clément

rproc_char_device_add will be called for each remoteproc that is added. So we will have one char dev per remoteproc.

Quoted 55 lines
Quoted 55 lines
drivers/remoteproc/remoteproc_userspace.c new file +19 -0
+
+	cdev_init(&rproc->char_dev, &rproc_fops);
+	rproc->char_dev.owner = THIS_MODULE;
+
+	cdevt = MKDEV(MAJOR(rproc_cdev), minor);
+	ret = cdev_add(&rproc->char_dev, cdevt, 1);
+	if (ret < 0)
+		ida_simple_remove(&cdev_minor_ida, minor);
+
+	rproc->dev.devt = cdevt;
+
+	return ret;
+}
+
+void __init rproc_init_cdev(void)
+{
+	int ret;
+
+	ret = alloc_chrdev_region(&rproc_cdev, 0, NUM_RPROC_DEVICES,

"rproc");

drivers/remoteproc/remoteproc_userspace.c new file +11 -0
+	if (ret < 0) {
+		pr_err("Failed to alloc rproc_cdev region, err %d\n", ret);
+		return;
+	}
+}
+
+void __exit rproc_exit_cdev(void)
+{
+	unregister_chrdev_region(MKDEV(MAJOR(rproc_cdev), 0),
+				NUM_RPROC_DEVICES);
+}
include/linux/remoteproc.h +2 -1
index 16ad666..c4ca796 100644
@@ -37,6 +37,7 @@
 
 #include <linux/types.h>
 #include <linux/mutex.h>
+#include <linux/cdev.h>
 #include <linux/virtio.h>
 #include <linux/completion.h>
 #include <linux/idr.h>
@@ -514,6 +515,7 @@ struct rproc {
 	bool auto_boot;
 	struct list_head dump_segments;
 	int nb_vdev;
+	struct cdev char_dev;
 };
 
 /**
--

The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project

maintainer
#6 lore raw

Re: [PATCH 1/2] remoteproc: Add userspace char device driver

In reply to Rishabh Bhatnagar
Quoted 8 lines · Rishabh Bhatnagar

Add the driver for creating the character device interface for userspace applications. The character device interface can be used in order to boot up and shutdown the remote processor. This might be helpful for remote processors that are booted by userspace applications and need to shutdown when the application crahes/shutsdown.

Change-Id: If23c8986272bb7c943eb76665f127ff706f12394

On my side checkpatch is complaining loudly that Change-Id tags should be removed... I wonder how you didn't get it on your side.

Quoted 53 lines
---
 drivers/remoteproc/Makefile               |  1 +
 drivers/remoteproc/remoteproc_internal.h  |  6 +++
 drivers/remoteproc/remoteproc_userspace.c | 90 +++++++++++++++++++++++++++++++
 include/linux/remoteproc.h                |  2 +
 4 files changed, 99 insertions(+)
 create mode 100644 drivers/remoteproc/remoteproc_userspace.c
drivers/remoteproc/Makefile +1 -0
index e30a1b1..facb3fa 100644
@@ -7,6 +7,7 @@ obj-$(CONFIG_REMOTEPROC)		+= remoteproc.o
 remoteproc-y				:= remoteproc_core.o
 remoteproc-y				+= remoteproc_debugfs.o
 remoteproc-y				+= remoteproc_sysfs.o
+remoteproc-y				+= remoteproc_userspace.o
 remoteproc-y				+= remoteproc_virtio.o
 remoteproc-y				+= remoteproc_elf_loader.o
 obj-$(CONFIG_IMX_REMOTEPROC)		+= imx_rproc.o
drivers/remoteproc/remoteproc_internal.h +6 -0
index 493ef92..97513ba 100644
@@ -47,6 +47,9 @@ struct dentry *rproc_create_trace_file(const char *name, struct rproc *rproc,
 int rproc_init_sysfs(void);
 void rproc_exit_sysfs(void);
 
+void rproc_init_cdev(void);
+void rproc_exit_cdev(void);
+
 void rproc_free_vring(struct rproc_vring *rvring);
 int rproc_alloc_vring(struct rproc_vdev *rvdev, int i);
 
@@ -63,6 +66,9 @@ struct resource_table *rproc_elf_find_loaded_rsc_table(struct rproc *rproc,
 struct rproc_mem_entry *
 rproc_find_carveout_by_name(struct rproc *rproc, const char *name, ...);
 
+/* from remoteproc_userspace.c */
+int rproc_char_device_add(struct rproc *rproc);
+
 static inline
 int rproc_fw_sanity_check(struct rproc *rproc, const struct firmware *fw)
 {
drivers/remoteproc/remoteproc_userspace.c new file +2 -0
new file mode 100644
index 0000000..2ef7679
@@ -0,0 +1,90 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*

drivers/remoteproc/remoteproc_userspace.c line 12

Quoted 10 lines
drivers/remoteproc/remoteproc_userspace.c new file +10 -0
+ * Character device interface driver for Remoteproc framework.
+ *
+ * Copyright (c) 2020, The Linux Foundation. All rights reserved.
+ */
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/cdev.h>
+#include <linux/mutex.h>
+#include <linux/remoteproc.h>

Alphabetical oder.

Quoted 10 lines
drivers/remoteproc/remoteproc_userspace.c new file +10 -0
+
+#include "remoteproc_internal.h"
+
+#define NUM_RPROC_DEVICES	64
+static dev_t rproc_cdev;
+static DEFINE_IDA(cdev_minor_ida);
+
+static int rproc_open(struct inode *inode, struct file *file)
+{
+	struct rproc *rproc;

drivers/remoteproc/remoteproc_userspace.c line 32

Quoted 10 lines
drivers/remoteproc/remoteproc_userspace.c new file +10 -0
+
+	rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+	if (!rproc)
+		return -EINVAL;
+
+	return rproc_boot(rproc);
+}
+
+static int rproc_release(struct inode *inode, struct file *file)
+{

This function declaration conflicts with the one already defined in remoteproc_internal.h. Again, I wonder how you didn't hit that problem when you compiled this patch.

drivers/remoteproc/remoteproc_userspace.c line 39

Quoted 7 lines
drivers/remoteproc/remoteproc_userspace.c new file +7 -0
+	struct rproc *rproc;
+
+	rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+	if (!rproc)
+		return -EINVAL;
+
+	rproc_shutdown(rproc);

The scenario I see here is that a userspace program will call open(/dev/rproc_xyz, SOME_OPTION) when it is launched. The file stays open until either the application shuts down, in which case it calls close() or it crashes. In that case the system will automatically close all file descriptors that were open by the application, which will also call rproc_shutdown().

To me the same functionality can be achieved with the current functionality provided by sysfs.

When the application starts it needs to read "/sys/class/remoteproc/remoteprocX/state". If the state is "offline" then "start" should be written to "/sys/.../state". If the state is "running" the application just crashed and got restarted. In which case it needs to stop the remote processor and start it again.

Quoted 6 lines
drivers/remoteproc/remoteproc_userspace.c new file +6 -0
+
+	return 0;
+}
+
+static const struct file_operations rproc_fops = {
+	.open = rproc_open,

drivers/remoteproc/remoteproc_userspace.c line 55

Quoted 10 lines
drivers/remoteproc/remoteproc_userspace.c new file +10 -0
+	.release = rproc_release,
+};
+
+int rproc_char_device_add(struct rproc *rproc)
+{
+	int ret, minor;
+	dev_t cdevt;
+
+	minor = ida_simple_get(&cdev_minor_ida, 0, NUM_RPROC_DEVICES,
+			GFP_KERNEL);

Indentation issue.

drivers/remoteproc/remoteproc_userspace.c line 57

Quoted 2 lines
drivers/remoteproc/remoteproc_userspace.c new file +2 -0
+	if (minor < 0) {
+	pr_err("%s: No more minor numbers left! rc:%d\n", __func__,

Here to...

drivers/remoteproc/remoteproc_userspace.c line 58

Quoted 1 line
drivers/remoteproc/remoteproc_userspace.c new file +1 -0
+							minor);

And here.

drivers/remoteproc/remoteproc_userspace.c line 68

Quoted 10 lines
drivers/remoteproc/remoteproc_userspace.c new file +10 -0
+		return -ENODEV;
+	}
+
+	cdev_init(&rproc->char_dev, &rproc_fops);
+	rproc->char_dev.owner = THIS_MODULE;
+
+	cdevt = MKDEV(MAJOR(rproc_cdev), minor);
+	ret = cdev_add(&rproc->char_dev, cdevt, 1);
+	if (ret < 0)
+		ida_simple_remove(&cdev_minor_ida, minor);

If the module is removed unregister_chrdev_region() is called but I don't see anywhere in there that cdev_del() is called.

Quoted 11 lines
drivers/remoteproc/remoteproc_userspace.c new file +11 -0
+
+	rproc->dev.devt = cdevt;
+
+	return ret;
+}
+
+void __init rproc_init_cdev(void)
+{
+	int ret;
+
+	ret = alloc_chrdev_region(&rproc_cdev, 0, NUM_RPROC_DEVICES, "rproc");

drivers/remoteproc/remoteproc_userspace.c line 89

Quoted 10 lines
drivers/remoteproc/remoteproc_userspace.c new file +10 -0
+	if (ret < 0) {
+		pr_err("Failed to alloc rproc_cdev region, err %d\n", ret);
+		return;
+	}
+}
+
+void __exit rproc_exit_cdev(void)
+{
+	unregister_chrdev_region(MKDEV(MAJOR(rproc_cdev), 0),
+				NUM_RPROC_DEVICES);

Indentation problem.

Quoted 14 lines
drivers/remoteproc/remoteproc_userspace.c new file +1 -0
+}
include/linux/remoteproc.h +1 -0
index 16ad666..c4ca796 100644
@@ -37,6 +37,7 @@
 
 #include <linux/types.h>
 #include <linux/mutex.h>
+#include <linux/cdev.h>
 #include <linux/virtio.h>
 #include <linux/completion.h>
 #include <linux/idr.h>

include/linux/remoteproc.h struct rproc { · line 518

Quoted 4 lines
include/linux/remoteproc.h +1 -0
@@ -514,6 +515,7 @@ struct rproc {
 	bool auto_boot;
 	struct list_head dump_segments;
 	int nb_vdev;
+	struct cdev char_dev;

The next time you send a patchset, please compile it and run checkpatch on it.

Thanks,
Mathieu

Quoted 6 lines
include/linux/remoteproc.h +0 -0
 };
 
 /**
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
maintainer
#12 lore raw

Re: [PATCH 2/2] remoteproc: core: Register the character device interface

In reply to Rishabh Bhatnagar
Quoted 31 lines · Rishabh Bhatnagar

Add the character device during rproc_add. This would create a character device node at /dev/subsys_<rproc_name>. Userspace applications can interact with the remote processor using this interface rather than using sysfs node. To distinguish between different remote processor nodes the device name has been changed to include the rproc name appended to "rproc_" string.

Change-Id I2114f77f8d2b5fd97e281021ec9905ef5c2fb54c
---
 drivers/remoteproc/remoteproc_core.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
drivers/remoteproc/remoteproc_core.c +6 -0
index 097f33e..f657983 100644
@@ -1907,6 +1907,12 @@ int rproc_add(struct rproc *rproc)
 	struct device *dev = &rproc->dev;
 	int ret;
 
+	ret = rproc_char_device_add(rproc);
+	if (ret) {
+		pr_err("error while adding character device\n");
+		return ret;
+	}
+
 	ret = device_add(dev);
 	if (ret < 0)
 		return ret;

drivers/remoteproc/remoteproc_core.c struct rproc *rproc_alloc(struct device *dev, const char *name, · line 2053

Quoted 5 lines
drivers/remoteproc/remoteproc_core.c +1 -1
@@ -2044,7 +2050,7 @@ struct rproc *rproc_alloc(struct device *dev, const char *name,
 		return NULL;
 	}
 
-	dev_set_name(&rproc->dev, "remoteproc%d", rproc->index);
+	dev_set_name(&rproc->dev, "rproc_%s", rproc->name);

Unfortunately you can't do that. The name of the remoteproc in sysfs is visible by users and a lot of scripts will fail because of this change.

Quoted 22 lines
drivers/remoteproc/remoteproc_core.c +2 -0
 	atomic_set(&rproc->power, 0);
 
@@ -2220,6 +2226,7 @@ static int __init remoteproc_init(void)
 {
 	rproc_init_sysfs();
 	rproc_init_debugfs();
+	rproc_init_cdev();
 
 	return 0;
 }
@@ -2231,6 +2238,7 @@ static void __exit remoteproc_exit(void)
 
 	rproc_exit_debugfs();
 	rproc_exit_sysfs();
+	rproc_exit_cdev();
 }
 module_exit(remoteproc_exit);
 
Signature · 2 lines
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project
#7 lore raw

Re: [PATCH 1/2] remoteproc: Add userspace char device driver

In reply to Mathieu Poirier

On Mon 30 Mar 15:12 PDT 2020, Mathieu Poirier wrote:
[..]

Quoted 14 lines

drivers/remoteproc/remoteproc_userspace.c line 39

Quoted 7 lines
drivers/remoteproc/remoteproc_userspace.c new file +7 -0
+	struct rproc *rproc;
+
+	rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+	if (!rproc)
+		return -EINVAL;
+
+	rproc_shutdown(rproc);

The scenario I see here is that a userspace program will call open(/dev/rproc_xyz, SOME_OPTION) when it is launched. The file stays open until either the application shuts down, in which case it calls close() or it crashes. In that case the system will automatically close all file descriptors that were open by the application, which will also call rproc_shutdown().

Quoted 9 lines

To me the same functionality can be achieved with the current functionality provided by sysfs.

When the application starts it needs to read "/sys/class/remoteproc/remoteprocX/state". If the state is "offline" then "start" should be written to "/sys/.../state". If the state is "running" the application just crashed and got restarted. In which case it needs to stop the remote processor and start it again.

A case when this would be useful is the Qualcomm modem, which relies on disk access through a "remote file system service" [1].

Today we register the service (a few layers ontop of rpmsg/GLINK) then find the modem remoteproc and write "start" into the state sysfs file.

When we get a signal for termination we write "stop" into state to stop the remoteproc before exiting.

There is however no way for us to indicate to the modem that rmtfs just died, e.g. a kill -9 on the process will result in the modem continue and the next IO request will fail which in most cases will be fatal.

So instead having rmtfs holding /dev/rproc_foo open would upon its termination cause the modem to be stopped automatically, and as the system respawns rmtfs the modem would be started anew and the two sides would be synced up again.

[1] https://github.com/andersson/rmtfs

Regards,
Bjorn

maintainer
#8 lore raw

Re: [PATCH 1/2] remoteproc: Add userspace char device driver

Quoted 33 lines · Bjorn Andersson

On Mon 30 Mar 15:12 PDT 2020, Mathieu Poirier wrote:
[..]

Quoted 14 lines

drivers/remoteproc/remoteproc_userspace.c line 39

Quoted 7 lines
drivers/remoteproc/remoteproc_userspace.c new file +7 -0
+   struct rproc *rproc;
+
+   rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+   if (!rproc)
+           return -EINVAL;
+
+   rproc_shutdown(rproc);

The scenario I see here is that a userspace program will call open(/dev/rproc_xyz, SOME_OPTION) when it is launched. The file stays open until either the application shuts down, in which case it calls close() or it crashes. In that case the system will automatically close all file descriptors that were open by the application, which will also call rproc_shutdown().

Quoted 9 lines

To me the same functionality can be achieved with the current functionality provided by sysfs.

When the application starts it needs to read "/sys/class/remoteproc/remoteprocX/state". If the state is "offline" then "start" should be written to "/sys/.../state". If the state is "running" the application just crashed and got restarted. In which case it needs to stop the remote processor and start it again.

A case when this would be useful is the Qualcomm modem, which relies on disk access through a "remote file system service" [1].

Today we register the service (a few layers ontop of rpmsg/GLINK) then find the modem remoteproc and write "start" into the state sysfs file.

Quoted 6 lines

When we get a signal for termination we write "stop" into state to stop the remoteproc before exiting.

There is however no way for us to indicate to the modem that rmtfs just died, e.g. a kill -9 on the process will result in the modem continue and the next IO request will fail which in most cases will be fatal.

The modem will crash when attempting an IO while rmtfs is down?

Quoted 5 lines

So instead having rmtfs holding /dev/rproc_foo open would upon its termination cause the modem to be stopped automatically, and as the system respawns rmtfs the modem would be started anew and the two sides would be synced up again.

I have a better idea of what is going on now - thanks for writing this up.

I would make this feature a kernel configurable option as some people may not want it. I also think having "/dev/remoteprocX" is fine, so no need to change anything currently visible in sysfs.

Mathieu

Quoted 5 lines

[1] https://github.com/andersson/rmtfs

Regards,
Bjorn

#9 lore raw

Re: [PATCH 1/2] remoteproc: Add userspace char device driver

Quoted 35 lines · Mathieu Poirier
Quoted 33 lines · Bjorn Andersson

On Mon 30 Mar 15:12 PDT 2020, Mathieu Poirier wrote:
[..]

Quoted 14 lines

drivers/remoteproc/remoteproc_userspace.c line 39

Quoted 7 lines
drivers/remoteproc/remoteproc_userspace.c new file +7 -0
+   struct rproc *rproc;
+
+   rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+   if (!rproc)
+           return -EINVAL;
+
+   rproc_shutdown(rproc);

The scenario I see here is that a userspace program will call open(/dev/rproc_xyz, SOME_OPTION) when it is launched. The file stays open until either the application shuts down, in which case it calls close() or it crashes. In that case the system will automatically close all file descriptors that were open by the application, which will also call rproc_shutdown().

Quoted 9 lines

To me the same functionality can be achieved with the current functionality provided by sysfs.

When the application starts it needs to read "/sys/class/remoteproc/remoteprocX/state". If the state is "offline" then "start" should be written to "/sys/.../state". If the state is "running" the application just crashed and got restarted. In which case it needs to stop the remote processor and start it again.

A case when this would be useful is the Qualcomm modem, which relies on disk access through a "remote file system service" [1].

Today we register the service (a few layers ontop of rpmsg/GLINK) then find the modem remoteproc and write "start" into the state sysfs file.

Quoted 6 lines
Quoted 6 lines

When we get a signal for termination we write "stop" into state to stop the remoteproc before exiting.

There is however no way for us to indicate to the modem that rmtfs just died, e.g. a kill -9 on the process will result in the modem continue and the next IO request will fail which in most cases will be fatal.

I have this scenario written down in the cover letter for this patchset
"[PATCH 0/2] Add character device interface to remoteproc"
I'll add it to the commit text as well.

Quoted 9 lines

The modem will crash when attempting an IO while rmtfs is down?

Quoted 5 lines

So instead having rmtfs holding /dev/rproc_foo open would upon its termination cause the modem to be stopped automatically, and as the system respawns rmtfs the modem would be started anew and the two sides would be synced up again.

Quoted 5 lines

I have a better idea of what is going on now - thanks for writing this up.

I would make this feature a kernel configurable option as some people may not want it. I also think having "/dev/remoteprocX" is fine, so no need to change anything currently visible in sysfs.

Ok. Makes sense.

Quoted 8 lines

Mathieu

Quoted 5 lines

[1] https://github.com/andersson/rmtfs

Regards,
Bjorn

#10 lore raw

Re: [PATCH 1/2] remoteproc: Add userspace char device driver

In reply to Mathieu Poirier
Quoted 35 lines · Mathieu Poirier
Quoted 33 lines · Bjorn Andersson

On Mon 30 Mar 15:12 PDT 2020, Mathieu Poirier wrote:
[..]

Quoted 14 lines

drivers/remoteproc/remoteproc_userspace.c line 39

Quoted 7 lines
drivers/remoteproc/remoteproc_userspace.c new file +7 -0
+   struct rproc *rproc;
+
+   rproc = container_of(inode->i_cdev, struct rproc, char_dev);
+   if (!rproc)
+           return -EINVAL;
+
+   rproc_shutdown(rproc);

The scenario I see here is that a userspace program will call open(/dev/rproc_xyz, SOME_OPTION) when it is launched. The file stays open until either the application shuts down, in which case it calls close() or it crashes. In that case the system will automatically close all file descriptors that were open by the application, which will also call rproc_shutdown().

Quoted 9 lines

To me the same functionality can be achieved with the current functionality provided by sysfs.

When the application starts it needs to read "/sys/class/remoteproc/remoteprocX/state". If the state is "offline" then "start" should be written to "/sys/.../state". If the state is "running" the application just crashed and got restarted. In which case it needs to stop the remote processor and start it again.

A case when this would be useful is the Qualcomm modem, which relies on disk access through a "remote file system service" [1].

Today we register the service (a few layers ontop of rpmsg/GLINK) then find the modem remoteproc and write "start" into the state sysfs file.

Quoted 9 lines
Quoted 6 lines

When we get a signal for termination we write "stop" into state to stop the remoteproc before exiting.

There is however no way for us to indicate to the modem that rmtfs just died, e.g. a kill -9 on the process will result in the modem continue and the next IO request will fail which in most cases will be fatal.

The modem will crash when attempting an IO while rmtfs is down?

In certain cases there's nothing else to do.

Quoted 10 lines
Quoted 5 lines

So instead having rmtfs holding /dev/rproc_foo open would upon its termination cause the modem to be stopped automatically, and as the system respawns rmtfs the modem would be started anew and the two sides would be synced up again.

I have a better idea of what is going on now - thanks for writing this up.

I would make this feature a kernel configurable option as some people may not want it.

Sounds reasonable.

Quoted 3 lines

I also think having "/dev/remoteprocX" is fine, so no need to change anything currently visible in sysfs.

I agree, it sure is annoying that remoteproc%d isn't stable, but it's what we have and consistency is important.

Regards,
Bjorn