Skip to contentSkip to conversation

vmcore: Add a kernel cmdline vmcore_device_dump

PATCH Superseded Kairui Song 3 participants 5 messages May 20, 2019 – May 23, 2019

Subsystems (5)

v2 of 5 Superseded Latest is v5 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 vmcore: Add a kernel cmdline vmcore_device_dump

Since commit 2724273e8fd0 ('vmcore: add API to collect hardware dump in second kernel'), drivers is allowed to add device related dump data to vmcore as they want by using the device dump API. This have a potential issue, the data is stored in memory, drivers may append too much data and use too much memory. The vmcore is typically used in a kdump kernel which runs in a pre-reserved small chunk of memory. So as a result it will make kdump unusable at all due to OOM issues.

So introduce new vmcore_device_dump= kernel parameter, and disable device dump by default. User can enable it only if device dump data is required for debugging, and have the chance to increase the kdump reserved memory accordingly before device dump fails kdump.

---
Update from V1:
- Use bool parameter to turn it on/off instead of letting user give the size limit. Size of device dump is hard to determine.

 Documentation/admin-guide/kernel-parameters.txt | 15 +++++++++++++++
 fs/proc/vmcore.c                                | 13 +++++++++++++
 2 files changed, 28 insertions(+)
Documentation/admin-guide/kernel-parameters.txt +15 -0
index 43176340c73d..2d48e39fd080 100644
@@ -5062,6 +5062,21 @@
 			decrease the size and leave more room for directly
 			mapped kernel RAM.
 
+	vmcore_device_dump=
+			[VMCORE]
+			Format: {"off" | "on"}
+			If CONFIG_PROC_VMCORE_DEVICE_DUMP is set,
+			this parameter allows enable or disable device dump
+			for vmcore.
+			Device dump allows drivers to append dump data to
+			vmcore so you can collect driver specified debug info.
+			Note that the drivers could append the data without
+			any limit, and the data is stored in memory, this may
+			bring a significant memory stress. If you want to turn
+			on this option, make sure you have reserved enough memory
+			with crashkernel= parameter.
+			default: off
+
 	vmcp_cma=nn[MG]	[KNL,S390]
 			Sets the memory size reserved for contiguous memory
 			allocations for the vmcp device driver.
fs/proc/vmcore.c +13 -0
index 3fe90443c1bb..d1b608b0efad 100644
@@ -53,6 +53,8 @@ static struct proc_dir_entry *proc_vmcore;
 /* Device Dump list and mutex to synchronize access to list */
 static LIST_HEAD(vmcoredd_list);
 static DEFINE_MUTEX(vmcoredd_mutex);
+
+static bool vmcoredd_enabled;
 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */
 
 /* Device Dump Size */
@@ -1451,6 +1453,11 @@ int vmcore_add_device_dump(struct vmcoredd_data *data)
 	size_t data_size;
 	int ret;
 
+	if (!vmcoredd_enabled) {
+		pr_err_once("Device dump is disabled\n");
+		return -EINVAL;
+	}
+
 	if (!data || !strlen(data->dump_name) ||
 	    !data->vmcoredd_callback || !data->size)
 		return -EINVAL;
@@ -1502,6 +1509,12 @@ int vmcore_add_device_dump(struct vmcoredd_data *data)
 	return ret;
 }
 EXPORT_SYMBOL(vmcore_add_device_dump);
+
+static int __init vmcoredd_parse_cmdline(char *arg)
+{
+	return kstrtobool(arg, &vmcoredd_enabled);
+}
+__setup("vmcore_device_dump=", vmcoredd_parse_cmdline);
 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */
 
 /* Free all dumps in vmcore device dump list */
Signature · 1 line
2.21.0
#1 lore raw
Quoted 28 lines · Kairui Song

Since commit 2724273e8fd0 ('vmcore: add API to collect hardware dump in second kernel'), drivers is allowed to add device related dump data to vmcore as they want by using the device dump API. This have a potential issue, the data is stored in memory, drivers may append too much data and use too much memory. The vmcore is typically used in a kdump kernel which runs in a pre-reserved small chunk of memory. So as a result it will make kdump unusable at all due to OOM issues.

So introduce new vmcore_device_dump= kernel parameter, and disable device dump by default. User can enable it only if device dump data is required for debugging, and have the chance to increase the kdump reserved memory accordingly before device dump fails kdump.

---
Update from V1:
- Use bool parameter to turn it on/off instead of letting user give the size limit. Size of device dump is hard to determine.

 Documentation/admin-guide/kernel-parameters.txt | 15 +++++++++++++++
 fs/proc/vmcore.c                                | 13 +++++++++++++
 2 files changed, 28 insertions(+)
Documentation/admin-guide/kernel-parameters.txt +0 -0
index 43176340c73d..2d48e39fd080 100644

Documentation/admin-guide/kernel-parameters.txt line 5066

Quoted 5 lines
Documentation/admin-guide/kernel-parameters.txt +2 -0
@@ -5062,6 +5062,21 @@
 			decrease the size and leave more room for directly
 			mapped kernel RAM.
 
+	vmcore_device_dump=
+			[VMCORE]

It looks better to have above two line merged in one line, also use [KNL, KDUMP] will be better.

Quoted 57 lines
Documentation/admin-guide/kernel-parameters.txt +13 -0
+			Format: {"off" | "on"}
+			If CONFIG_PROC_VMCORE_DEVICE_DUMP is set,
+			this parameter allows enable or disable device dump
+			for vmcore.
+			Device dump allows drivers to append dump data to
+			vmcore so you can collect driver specified debug info.
+			Note that the drivers could append the data without
+			any limit, and the data is stored in memory, this may
+			bring a significant memory stress. If you want to turn
+			on this option, make sure you have reserved enough memory
+			with crashkernel= parameter.
+			default: off
+
 	vmcp_cma=nn[MG]	[KNL,S390]
 			Sets the memory size reserved for contiguous memory
 			allocations for the vmcp device driver.
fs/proc/vmcore.c +13 -0
index 3fe90443c1bb..d1b608b0efad 100644
@@ -53,6 +53,8 @@ static struct proc_dir_entry *proc_vmcore;
 /* Device Dump list and mutex to synchronize access to list */
 static LIST_HEAD(vmcoredd_list);
 static DEFINE_MUTEX(vmcoredd_mutex);
+
+static bool vmcoredd_enabled;
 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */
 
 /* Device Dump Size */
@@ -1451,6 +1453,11 @@ int vmcore_add_device_dump(struct vmcoredd_data *data)
 	size_t data_size;
 	int ret;
 
+	if (!vmcoredd_enabled) {
+		pr_err_once("Device dump is disabled\n");
+		return -EINVAL;
+	}
+
 	if (!data || !strlen(data->dump_name) ||
 	    !data->vmcoredd_callback || !data->size)
 		return -EINVAL;
@@ -1502,6 +1509,12 @@ int vmcore_add_device_dump(struct vmcoredd_data *data)
 	return ret;
 }
 EXPORT_SYMBOL(vmcore_add_device_dump);
+
+static int __init vmcoredd_parse_cmdline(char *arg)
+{
+	return kstrtobool(arg, &vmcoredd_enabled);
+}
+__setup("vmcore_device_dump=", vmcoredd_parse_cmdline);
 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */
 
 /* Free all dumps in vmcore device dump list */
Signature · 1 line
2.21.0

Thanks
Dave

#2 lore raw
Quoted 92 lines · Dave Young
Quoted 28 lines · Kairui Song

Since commit 2724273e8fd0 ('vmcore: add API to collect hardware dump in second kernel'), drivers is allowed to add device related dump data to vmcore as they want by using the device dump API. This have a potential issue, the data is stored in memory, drivers may append too much data and use too much memory. The vmcore is typically used in a kdump kernel which runs in a pre-reserved small chunk of memory. So as a result it will make kdump unusable at all due to OOM issues.

So introduce new vmcore_device_dump= kernel parameter, and disable device dump by default. User can enable it only if device dump data is required for debugging, and have the chance to increase the kdump reserved memory accordingly before device dump fails kdump.

---
Update from V1:
- Use bool parameter to turn it on/off instead of letting user give the size limit. Size of device dump is hard to determine.

 Documentation/admin-guide/kernel-parameters.txt | 15 +++++++++++++++
 fs/proc/vmcore.c                                | 13 +++++++++++++
 2 files changed, 28 insertions(+)
Documentation/admin-guide/kernel-parameters.txt +0 -0
index 43176340c73d..2d48e39fd080 100644

Documentation/admin-guide/kernel-parameters.txt line 5066

Quoted 5 lines
Documentation/admin-guide/kernel-parameters.txt +2 -0
@@ -5062,6 +5062,21 @@
                      decrease the size and leave more room for directly
                      mapped kernel RAM.
 
+     vmcore_device_dump=
+                     [VMCORE]

It looks better to have above two line merged in one line, also use [KNL, KDUMP] will be better.

Quoted 53 lines
Documentation/admin-guide/kernel-parameters.txt +13 -0
+                     Format: {"off" | "on"}
+                     If CONFIG_PROC_VMCORE_DEVICE_DUMP is set,
+                     this parameter allows enable or disable device dump
+                     for vmcore.
+                     Device dump allows drivers to append dump data to
+                     vmcore so you can collect driver specified debug info.
+                     Note that the drivers could append the data without
+                     any limit, and the data is stored in memory, this may
+                     bring a significant memory stress. If you want to turn
+                     on this option, make sure you have reserved enough memory
+                     with crashkernel= parameter.
+                     default: off
+
      vmcp_cma=nn[MG] [KNL,S390]
                      Sets the memory size reserved for contiguous memory
                      allocations for the vmcp device driver.
fs/proc/vmcore.c +13 -0
index 3fe90443c1bb..d1b608b0efad 100644
@@ -53,6 +53,8 @@ static struct proc_dir_entry *proc_vmcore;
 /* Device Dump list and mutex to synchronize access to list */
 static LIST_HEAD(vmcoredd_list);
 static DEFINE_MUTEX(vmcoredd_mutex);
+
+static bool vmcoredd_enabled;
 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */
 
 /* Device Dump Size */
@@ -1451,6 +1453,11 @@ int vmcore_add_device_dump(struct vmcoredd_data *data)
      size_t data_size;
      int ret;
 
+     if (!vmcoredd_enabled) {
+             pr_err_once("Device dump is disabled\n");
+             return -EINVAL;
+     }
+
      if (!data || !strlen(data->dump_name) ||
          !data->vmcoredd_callback || !data->size)
              return -EINVAL;
@@ -1502,6 +1509,12 @@ int vmcore_add_device_dump(struct vmcoredd_data *data)
      return ret;
 }
 EXPORT_SYMBOL(vmcore_add_device_dump);
+
+static int __init vmcoredd_parse_cmdline(char *arg)
+{
+     return kstrtobool(arg, &vmcoredd_enabled);
+}
+__setup("vmcore_device_dump=", vmcoredd_parse_cmdline);
 #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */
 
Quoted 7 lines
Quoted 4 lines
fs/proc/vmcore.c +0 -1
 /* Free all dumps in vmcore device dump list */
--

2.21.0

Thanks
Dave

Good suggestion, I'll update in V3.

Signature · 7 lines
Best Regards,
Kairui Song

_______________________________________________
kexec mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/kexec
#3 lore raw
In reply to Kairui Song
Quoted 28 lines · Kairui Song

Since commit 2724273e8fd0 ('vmcore: add API to collect hardware dump in second kernel'), drivers is allowed to add device related dump data to vmcore as they want by using the device dump API. This have a potential issue, the data is stored in memory, drivers may append too much data and use too much memory. The vmcore is typically used in a kdump kernel which runs in a pre-reserved small chunk of memory. So as a result it will make kdump unusable at all due to OOM issues.

So introduce new vmcore_device_dump= kernel parameter, and disable device dump by default. User can enable it only if device dump data is required for debugging, and have the chance to increase the kdump reserved memory accordingly before device dump fails kdump.

---
Update from V1:
- Use bool parameter to turn it on/off instead of letting user give the size limit. Size of device dump is hard to determine.

  Documentation/admin-guide/kernel-parameters.txt | 15 +++++++++++++++
  fs/proc/vmcore.c                                | 13 +++++++++++++
  2 files changed, 28 insertions(+)
Documentation/admin-guide/kernel-parameters.txt +0 -0
index 43176340c73d..2d48e39fd080 100644

Documentation/admin-guide/kernel-parameters.txt line 5070

Quoted 9 lines
Documentation/admin-guide/kernel-parameters.txt +6 -0
@@ -5062,6 +5062,21 @@
  			decrease the size and leave more room for directly
  			mapped kernel RAM.
 
+	vmcore_device_dump=
+			[VMCORE]
+			Format: {"off" | "on"}
+			If CONFIG_PROC_VMCORE_DEVICE_DUMP is set,
+			this parameter allows enable or disable device dump
+			for vmcore.

We can add a simpler description here, something like:
Depends on CONFIG_PROC_VMCORE_DEVICE_DUMP

Documentation/admin-guide/kernel-parameters.txt line 5078

Quoted 8 lines
Documentation/admin-guide/kernel-parameters.txt +8 -0
+			Device dump allows drivers to append dump data to
+			vmcore so you can collect driver specified debug info.
+			Note that the drivers could append the data without
+			any limit, and the data is stored in memory, this may
+			bring a significant memory stress. If you want to turn
+			on this option, make sure you have reserved enough memory
+			with crashkernel= parameter.
+			default: off

... and massage the rest of text accordingly.

Better to also modify the help text for 'PROC_VMCORE_DEVICE_DUMP' config option defined in 'fs/proc/Kconfig'. Something like:

config PROC_VMCORE_DEVICE_DUMP
bool "Device Hardware/Firmware Log Collection"
<..snip..>
If you say Y here, the collected device dumps will be added as ELF notes to /proc/vmcore.

If this option is selected, device dump collection can still be
disabled by passing vmcore_device_dump=off to the kernel.

See config INTEL_IOMMU_DEFAULT_ON in 'drivers/iommu/Kconfig' as an example.

Quoted 32 lines
Documentation/admin-guide/kernel-parameters.txt +0 -0
  	vmcp_cma=nn[MG]	[KNL,S390]
  			Sets the memory size reserved for contiguous memory
  			allocations for the vmcp device driver.
fs/proc/vmcore.c +7 -0
index 3fe90443c1bb..d1b608b0efad 100644
@@ -53,6 +53,8 @@ static struct proc_dir_entry *proc_vmcore;
  /* Device Dump list and mutex to synchronize access to list */
  static LIST_HEAD(vmcoredd_list);
  static DEFINE_MUTEX(vmcoredd_mutex);
+
+static bool vmcoredd_enabled;
  #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */
 
  /* Device Dump Size */
@@ -1451,6 +1453,11 @@ int vmcore_add_device_dump(struct vmcoredd_data *data)
  	size_t data_size;
  	int ret;
 
+	if (!vmcoredd_enabled) {
+		pr_err_once("Device dump is disabled\n");
+		return -EINVAL;
+	}
+
  	if (!data || !strlen(data->dump_name) ||
  	    !data->vmcoredd_callback || !data->size)
  		return -EINVAL;
@@ -1502,6 +1509,12 @@ int vmcore_add_device_dump(struct vmcoredd_data *data)
  	return ret;
  }

fs/proc/vmcore.c int vmcore_add_device_dump(struct vmcoredd_data *data) · line 1520

Quoted 10 lines
fs/proc/vmcore.c +6 -0
  EXPORT_SYMBOL(vmcore_add_device_dump);
+
+static int __init vmcoredd_parse_cmdline(char *arg)
+{
+	return kstrtobool(arg, &vmcoredd_enabled);
+}
+__setup("vmcore_device_dump=", vmcoredd_parse_cmdline);
  #endif /* CONFIG_PROC_VMCORE_DEVICE_DUMP */
 
  /* Free all dumps in vmcore device dump list */

Thanks,
Bhupesh

#4 lore raw
Quoted 63 lines · Bhupesh Sharma
Quoted 28 lines · Kairui Song

Since commit 2724273e8fd0 ('vmcore: add API to collect hardware dump in second kernel'), drivers is allowed to add device related dump data to vmcore as they want by using the device dump API. This have a potential issue, the data is stored in memory, drivers may append too much data and use too much memory. The vmcore is typically used in a kdump kernel which runs in a pre-reserved small chunk of memory. So as a result it will make kdump unusable at all due to OOM issues.

So introduce new vmcore_device_dump= kernel parameter, and disable device dump by default. User can enable it only if device dump data is required for debugging, and have the chance to increase the kdump reserved memory accordingly before device dump fails kdump.

---
Update from V1:
- Use bool parameter to turn it on/off instead of letting user give the size limit. Size of device dump is hard to determine.

  Documentation/admin-guide/kernel-parameters.txt | 15 +++++++++++++++
  fs/proc/vmcore.c                                | 13 +++++++++++++
  2 files changed, 28 insertions(+)
Documentation/admin-guide/kernel-parameters.txt +0 -0
index 43176340c73d..2d48e39fd080 100644

Documentation/admin-guide/kernel-parameters.txt line 5070

Quoted 9 lines
Documentation/admin-guide/kernel-parameters.txt +6 -0
@@ -5062,6 +5062,21 @@
                      decrease the size and leave more room for directly
                      mapped kernel RAM.
 
+     vmcore_device_dump=
+                     [VMCORE]
+                     Format: {"off" | "on"}
+                     If CONFIG_PROC_VMCORE_DEVICE_DUMP is set,
+                     this parameter allows enable or disable device dump
+                     for vmcore.

We can add a simpler description here, something like:
Depends on CONFIG_PROC_VMCORE_DEVICE_DUMP

Documentation/admin-guide/kernel-parameters.txt line 5078

Quoted 8 lines
Documentation/admin-guide/kernel-parameters.txt +8 -0
+                     Device dump allows drivers to append dump data to
+                     vmcore so you can collect driver specified debug info.
+                     Note that the drivers could append the data without
+                     any limit, and the data is stored in memory, this may
+                     bring a significant memory stress. If you want to turn
+                     on this option, make sure you have reserved enough memory
+                     with crashkernel= parameter.
+                     default: off

... and massage the rest of text accordingly.

Better to also modify the help text for 'PROC_VMCORE_DEVICE_DUMP' config option defined in 'fs/proc/Kconfig'. Something like:

config PROC_VMCORE_DEVICE_DUMP
bool "Device Hardware/Firmware Log Collection"
<..snip..>
If you say Y here, the collected device dumps will be added as ELF notes to /proc/vmcore.

Quoted 5 lines

If this option is selected, device dump collection can still be
disabled by passing vmcore_device_dump=off to the kernel.

See config INTEL_IOMMU_DEFAULT_ON in 'drivers/iommu/Kconfig' as an example.

Good suggestion! I'll update in V3.

Signature · 2 lines
Best Regards,
Kairui Song