[PATCH v3 1/4] remoteproc: Add remote processor coredump support

[Bjorn Andersson]

From: Sarangdhar Joshi <spjoshi@xxxxxxxxxxxxxx>

As the remoteproc framework restarts the remote processor after a fatal
event, it's useful to be able to acquire a coredump of the remote
processor's state, for post mortem debugging.

This patch introduces a mechanism for extracting the memory contents
after the remote has stopped and before the restart sequence has begun
in the recovery path. The remoteproc framework builds the core dump in
memory and use devcoredump to expose this to user space.

Signed-off-by: Sarangdhar Joshi <spjoshi@xxxxxxxxxxxxxx>
[bjorn: Use vmalloc instead of composing the ELF on the fly]
Signed-off-by: Bjorn Andersson <bjorn.andersson@xxxxxxxxxx>
---

Changes since v2:
- None

Changes since v1:
- vmalloc() memory to build ELF file before registering devcoredump
- life cycle of segment list is from boot to shutdown (not start/shutdown)
- add WANT_DEV_COREDUMP select
- use dma_addr_t to denote device addresses

 drivers/remoteproc/Kconfig           |   1 +
 drivers/remoteproc/remoteproc_core.c | 128 +++++++++++++++++++++++++++++++++++
 include/linux/remoteproc.h           |  18 +++++
 3 files changed, 147 insertions(+)

diff --git a/drivers/remoteproc/Kconfig b/drivers/remoteproc/Kconfig
index b609e1d3654b..3e4bca77188d 100644
--- a/drivers/remoteproc/Kconfig
+++ b/drivers/remoteproc/Kconfig
@@ -6,6 +6,7 @@ config REMOTEPROC
 	select CRC32
 	select FW_LOADER
 	select VIRTIO
+	select WANT_DEV_COREDUMP
 	help
 	  Support for remote processors (such as DSP coprocessors). These
 	  are mainly used on embedded systems.
diff --git a/drivers/remoteproc/remoteproc_core.c b/drivers/remoteproc/remoteproc_core.c
index 4170dfbd93bd..5af7547b9d8d 100644
--- a/drivers/remoteproc/remoteproc_core.c
+++ b/drivers/remoteproc/remoteproc_core.c
@@ -33,6 +33,7 @@
 #include <linux/firmware.h>
 #include <linux/string.h>
 #include <linux/debugfs.h>
+#include <linux/devcoredump.h>
 #include <linux/remoteproc.h>
 #include <linux/iommu.h>
 #include <linux/idr.h>
@@ -801,6 +802,20 @@ static void rproc_remove_subdevices(struct rproc *rproc)
 		subdev->remove(subdev);
 }
 
+/**
+ * rproc_coredump_cleanup() - clean up dump_segments list
+ * @rproc: the remote processor handle
+ */
+static void rproc_coredump_cleanup(struct rproc *rproc)
+{
+	struct rproc_dump_segment *entry, *tmp;
+
+	list_for_each_entry_safe(entry, tmp, &rproc->dump_segments, node) {
+		list_del(&entry->node);
+		kfree(entry);
+	}
+}
+
 /**
  * rproc_resource_cleanup() - clean up and free all acquired resources
  * @rproc: rproc handle
@@ -848,6 +863,8 @@ static void rproc_resource_cleanup(struct rproc *rproc)
 	/* clean up remote vdev entries */
 	list_for_each_entry_safe(rvdev, rvtmp, &rproc->rvdevs, node)
 		kref_put(&rvdev->refcount, rproc_vdev_release);
+
+	rproc_coredump_cleanup(rproc);
 }
 
 static int rproc_start(struct rproc *rproc, const struct firmware *fw)
@@ -1017,6 +1034,113 @@ static int rproc_stop(struct rproc *rproc)
 	return 0;
 }
 
+/**
+ * rproc_coredump_add_segment() - add segment of device memory to coredump
+ * @rproc:	handle of a remote processor
+ * @da:		device address
+ * @size:	size of segment
+ *
+ * Add device memory to the list of segments to be included in a coredump for
+ * the remoteproc.
+ *
+ * Return: 0 on success, negative errno on error.
+ */
+int rproc_coredump_add_segment(struct rproc *rproc, dma_addr_t da, size_t size)
+{
+	struct rproc_dump_segment *segment;
+
+	segment = kzalloc(sizeof(*segment), GFP_KERNEL);
+	if (!segment)
+		return -ENOMEM;
+
+	segment->da = da;
+	segment->size = size;
+
+	list_add_tail(&segment->node, &rproc->dump_segments);
+
+	return 0;
+}
+EXPORT_SYMBOL(rproc_coredump_add_segment);
+
+/**
+ * rproc_coredump() - perform coredump
+ * @rproc:	rproc handle
+ *
+ * This function will generate an ELF header for the registered segments
+ * and create a devcoredump device associated with rproc.
+ */
+static void rproc_coredump(struct rproc *rproc)
+{
+	struct rproc_dump_segment *segment;
+	struct elf32_phdr *phdr;
+	struct elf32_hdr *ehdr;
+	size_t data_size;
+	size_t offset;
+	void *data;
+	void *ptr;
+	int phnum = 0;
+
+	if (list_empty(&rproc->dump_segments))
+		return;
+
+	data_size = sizeof(*ehdr);
+	list_for_each_entry(segment, &rproc->dump_segments, node) {
+		data_size += sizeof(*phdr) + segment->size;
+
+		phnum++;
+	}
+
+	data = vmalloc(data_size);
+	if (!data)
+		return;
+
+	ehdr = data;
+
+	memset(ehdr, 0, sizeof(*ehdr));
+	memcpy(ehdr->e_ident, ELFMAG, SELFMAG);
+	ehdr->e_ident[EI_CLASS] = ELFCLASS32;
+	ehdr->e_ident[EI_DATA] = ELFDATA2LSB;
+	ehdr->e_ident[EI_VERSION] = EV_CURRENT;
+	ehdr->e_ident[EI_OSABI] = ELFOSABI_NONE;
+	ehdr->e_type = ET_CORE;
+	ehdr->e_machine = EM_NONE;
+	ehdr->e_version = EV_CURRENT;
+	ehdr->e_entry = rproc->bootaddr;
+	ehdr->e_phoff = sizeof(*ehdr);
+	ehdr->e_ehsize = sizeof(*ehdr);
+	ehdr->e_phentsize = sizeof(*phdr);
+	ehdr->e_phnum = phnum;
+
+	phdr = data + ehdr->e_phoff;
+	offset = ehdr->e_phoff + sizeof(*phdr) * ehdr->e_phnum;
+	list_for_each_entry(segment, &rproc->dump_segments, node) {
+		memset(phdr, 0, sizeof(*phdr));
+		phdr->p_type = PT_LOAD;
+		phdr->p_offset = offset;
+		phdr->p_vaddr = segment->da;
+		phdr->p_paddr = segment->da;
+		phdr->p_filesz = segment->size;
+		phdr->p_memsz = segment->size;
+		phdr->p_flags = PF_R | PF_W | PF_X;
+		phdr->p_align = 0;
+
+		ptr = rproc_da_to_va(rproc, segment->da, segment->size);
+		if (!ptr) {
+			dev_err(&rproc->dev,
+				"invalid coredump segment (%pad, %zu)\n",
+				&segment->da, segment->size);
+			memset(data + offset, 0xff, segment->size);
+		} else {
+			memcpy(data + offset, ptr, segment->size);
+		}
+
+		offset += phdr->p_filesz;
+		phdr++;
+	}
+
+	dev_coredumpv(&rproc->dev, data, data_size, GFP_KERNEL);
+}
+
 /**
  * rproc_trigger_recovery() - recover a remoteproc
  * @rproc: the remote processor
@@ -1043,6 +1167,9 @@ int rproc_trigger_recovery(struct rproc *rproc)
 	if (ret)
 		goto unlock_mutex;
 
+	/* generate coredump */
+	rproc_coredump(rproc);
+
 	/* load firmware */
 	ret = request_firmware(&firmware_p, rproc->firmware, dev);
 	if (ret < 0) {
@@ -1443,6 +1570,7 @@ struct rproc *rproc_alloc(struct device *dev, const char *name,
 	INIT_LIST_HEAD(&rproc->traces);
 	INIT_LIST_HEAD(&rproc->rvdevs);
 	INIT_LIST_HEAD(&rproc->subdevs);
+	INIT_LIST_HEAD(&rproc->dump_segments);
 
 	INIT_WORK(&rproc->crash_handler, rproc_crash_handler_work);
 
diff --git a/include/linux/remoteproc.h b/include/linux/remoteproc.h
index 728d421fffe9..b60c3a31b75d 100644
--- a/include/linux/remoteproc.h
+++ b/include/linux/remoteproc.h
@@ -394,6 +394,21 @@ enum rproc_crash_type {
 	RPROC_FATAL_ERROR,
 };
 
+/**
+ * struct rproc_dump_segment - segment info from ELF header
+ * @node:	list node related to the rproc segment list
+ * @da:		device address of the segment
+ * @size:	size of the segment
+ */
+struct rproc_dump_segment {
+	struct list_head node;
+
+	dma_addr_t da;
+	size_t size;
+
+	loff_t offset;
+};
+
 /**
  * struct rproc - represents a physical remote processor device
  * @node: list node of this rproc object
@@ -424,6 +439,7 @@ enum rproc_crash_type {
  * @cached_table: copy of the resource table
  * @table_sz: size of @cached_table
  * @has_iommu: flag to indicate if remote processor is behind an MMU
+ * @dump_segments: list of segments in the firmware
  */
 struct rproc {
 	struct list_head node;
@@ -455,6 +471,7 @@ struct rproc {
 	size_t table_sz;
 	bool has_iommu;
 	bool auto_boot;
+	struct list_head dump_segments;
 };
 
 /**
@@ -534,6 +551,7 @@ void rproc_free(struct rproc *rproc);
 int rproc_boot(struct rproc *rproc);
 void rproc_shutdown(struct rproc *rproc);
 void rproc_report_crash(struct rproc *rproc, enum rproc_crash_type type);
+int rproc_coredump_add_segment(struct rproc *rproc, dma_addr_t da, size_t size);
 
 static inline struct rproc_vdev *vdev_to_rvdev(struct virtio_device *vdev)
 {
-- 
2.15.0