Skip to contentSkip to conversation

kthread: dynamically allocate memory to store kthread's full name

PATCH Applied Yafang Shao 3 participants 5 messages Nov 20, 2021 – Nov 25, 2021

Subsystems (6)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 kthread: dynamically allocate memory to store kthread's full name

When I was implementing a new per-cpu kthread cfs_migration, I found the comm of it "cfs_migration/%u" is truncated due to the limitation of TASK_COMM_LEN. For example, the comm of the percpu thread on CPU10~19 are all with the same name "cfs_migration/1", which will confuse the user. This issue is not critical, because we can get the corresponding CPU from the task's Cpus_allowed. But for kthreads correspoinding to other hardware devices, it is not easy to get the detailed device info from task comm, for example,

    jbd2/nvme0n1p2-
    xfs-reclaim/sdf

Currently there are so many truncated kthreads:

    rcu_tasks_kthre
    rcu_tasks_rude_
    rcu_tasks_trace
    poll_mpt3sas0_s
    ext4-rsv-conver
    xfs-reclaim/sd{a, b, c, ...}
    xfs-blockgc/sd{a, b, c, ...}
    xfs-inodegc/sd{a, b, c, ...}
    audit_send_repl
    ecryptfs-kthrea
    vfio-irqfd-clea
    jbd2/nvme0n1p2-
    ...

We can shorten these names to work around this problem, but it may be not applied to all of the truncated kthreads. Take 'jbd2/nvme0n1p2-' for example, it is a nice name, and it is not a good idea to shorten it.

One possible way to fix this issue is extending the task comm size, but as task->comm is used in lots of places, that may cause some potential buffer overflows. Another more conservative approach is introducing a new pointer to store kthread's full name if it is truncated, which won't introduce too much overhead as it is in the non-critical path. Finally we make a dicision to use the second approach. See also the discussions in this thread:
https://lore.kernel.org/lkml/[email protected]/

After this change, the full name of these truncated kthreads will be displayed via /proc/[pid]/comm:

    rcu_tasks_kthread
    rcu_tasks_rude_kthread
    rcu_tasks_trace_kthread
    poll_mpt3sas0_statu
    ext4-rsv-conversion
    xfs-reclaim/sdf1
    xfs-blockgc/sdf1
    xfs-inodegc/sdf1
    audit_send_reply
    ecryptfs-kthread
    vfio-irqfd-cleanup
    jbd2/nvme0n1p2-8
Cc Mathieu Desnoyers <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Alexei Starovoitov <[email protected]>
Cc Andrii Nakryiko <[email protected]>
Cc Michal Miroslaw <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Steven Rostedt <[email protected]>
Cc Matthew Wilcox <[email protected]>
Cc David Hildenbrand <[email protected]>
Cc Al Viro <[email protected]>
Cc Kees Cook <[email protected]>
Cc Petr Mladek <[email protected]>

---

Changes since v1:
1. leave it turncated when out of memory (Kees & Petr)
2. do null check in free_kthread_struct (Petr)

---
 fs/proc/array.c         |  3 +++
 include/linux/kthread.h |  1 +
 kernel/kthread.c        | 32 ++++++++++++++++++++++++++++++--
 3 files changed, 34 insertions(+), 2 deletions(-)
fs/proc/array.c +3 -0
index ff869a66b34e..4321aa63835d 100644
@@ -92,6 +92,7 @@
 #include <linux/string_helpers.h>
 #include <linux/user_namespace.h>
 #include <linux/fs_struct.h>
+#include <linux/kthread.h>
 
 #include <asm/processor.h>
 #include "internal.h"
@@ -102,6 +103,8 @@ void proc_task_name(struct seq_file *m, struct task_struct *p, bool escape)
 
 	if (p->flags & PF_WQ_WORKER)
 		wq_worker_comm(tcomm, sizeof(tcomm), p);
+	else if (p->flags & PF_KTHREAD)
+		get_kthread_comm(tcomm, sizeof(tcomm), p);
 	else
 		__get_task_comm(tcomm, sizeof(tcomm), p);
 
include/linux/kthread.h +1 -0
index 346b0f269161..2a5c04494663 100644
@@ -33,6 +33,7 @@ struct task_struct *kthread_create_on_cpu(int (*threadfn)(void *data),
 					  unsigned int cpu,
 					  const char *namefmt);
 
+void get_kthread_comm(char *buf, size_t buf_size, struct task_struct *tsk);
 void set_kthread_struct(struct task_struct *p);
 
 void kthread_set_per_cpu(struct task_struct *k, int cpu);
kernel/kthread.c +30 -2
index 7113003fab63..a70cd5dc94e3 100644
@@ -60,6 +60,8 @@ struct kthread {
 #ifdef CONFIG_BLK_CGROUP
 	struct cgroup_subsys_state *blkcg_css;
 #endif
+	/* To store the full name if task comm is truncated. */
+	char *full_name;
 };
 
 enum KTHREAD_BITS {
@@ -93,6 +95,18 @@ static inline struct kthread *__to_kthread(struct task_struct *p)
 	return kthread;
 }
 
+void get_kthread_comm(char *buf, size_t buf_size, struct task_struct *tsk)
+{
+	struct kthread *kthread = to_kthread(tsk);
+
+	if (!kthread || !kthread->full_name) {
+		__get_task_comm(buf, buf_size, tsk);
+		return;
+	}
+
+	strscpy_pad(buf, kthread->full_name, buf_size);
+}
+
 void set_kthread_struct(struct task_struct *p)
 {
 	struct kthread *kthread;
@@ -118,9 +132,13 @@ void free_kthread_struct(struct task_struct *k)
 	 * or if kmalloc() in kthread() failed.
 	 */
 	kthread = to_kthread(k);
+	if (!kthread)
+		return;
+
 #ifdef CONFIG_BLK_CGROUP
-	WARN_ON_ONCE(kthread && kthread->blkcg_css);
+	WARN_ON_ONCE(kthread->blkcg_css);
 #endif
+	kfree(kthread->full_name);
 	kfree(kthread);
 }
 
@@ -406,12 +424,22 @@ struct task_struct *__kthread_create_on_node(int (*threadfn)(void *data),
 	task = create->result;
 	if (!IS_ERR(task)) {
 		char name[TASK_COMM_LEN];
+		va_list aq;
+		int len;
 
 		/*
 		 * task is already visible to other tasks, so updating
 		 * COMM must be protected.
 		 */
-		vsnprintf(name, sizeof(name), namefmt, args);
+		va_copy(aq, args);
+		len = vsnprintf(name, sizeof(name), namefmt, aq);
+		va_end(aq);
+		if (len >= TASK_COMM_LEN) {
+			struct kthread *kthread = to_kthread(task);
+
+			/* leave it truncated when out of memory. */
+			kthread->full_name = kvasprintf(GFP_KERNEL, namefmt, args);
+		}
 		set_task_comm(task, name);
 	}
 	kfree(create);
Signature · 1 line
2.17.1
Reviewed-by on v2 1/1
#1 lore raw
Quoted 13 lines · Yafang Shao

When I was implementing a new per-cpu kthread cfs_migration, I found the comm of it "cfs_migration/%u" is truncated due to the limitation of TASK_COMM_LEN.

One possible way to fix this issue is extending the task comm size, but as task->comm is used in lots of places, that may cause some potential buffer overflows. Another more conservative approach is introducing a new pointer to store kthread's full name if it is truncated, which won't introduce too much overhead as it is in the non-critical path. Finally we make a dicision to use the second approach. See also the discussions in this thread:
https://lore.kernel.org/lkml/[email protected]/

Quoted 4 lines

After this change, the full name of these truncated kthreads will be displayed via /proc/[pid]/comm:

Looks good to me:

Best Regards,
Petr

Reviewed-by on v2 1/1
#2 lore raw
In reply to Yafang Shao
Quoted 44 lines · Yafang Shao

When I was implementing a new per-cpu kthread cfs_migration, I found the comm of it "cfs_migration/%u" is truncated due to the limitation of TASK_COMM_LEN. For example, the comm of the percpu thread on CPU10~19 are all with the same name "cfs_migration/1", which will confuse the user. This issue is not critical, because we can get the corresponding CPU from the task's Cpus_allowed. But for kthreads correspoinding to other hardware devices, it is not easy to get the detailed device info from task comm, for example,

    jbd2/nvme0n1p2-
    xfs-reclaim/sdf

Currently there are so many truncated kthreads:

    rcu_tasks_kthre
    rcu_tasks_rude_
    rcu_tasks_trace
    poll_mpt3sas0_s
    ext4-rsv-conver
    xfs-reclaim/sd{a, b, c, ...}
    xfs-blockgc/sd{a, b, c, ...}
    xfs-inodegc/sd{a, b, c, ...}
    audit_send_repl
    ecryptfs-kthrea
    vfio-irqfd-clea
    jbd2/nvme0n1p2-
    ...

We can shorten these names to work around this problem, but it may be not applied to all of the truncated kthreads. Take 'jbd2/nvme0n1p2-' for example, it is a nice name, and it is not a good idea to shorten it.

One possible way to fix this issue is extending the task comm size, but as task->comm is used in lots of places, that may cause some potential buffer overflows. Another more conservative approach is introducing a new pointer to store kthread's full name if it is truncated, which won't introduce too much overhead as it is in the non-critical path. Finally we make a dicision to use the second approach. See also the discussions in this thread:
https://lore.kernel.org/lkml/[email protected]/

After this change, the full name of these truncated kthreads will be displayed via /proc/[pid]/comm:

Quoted 12 lines
    rcu_tasks_kthread
    rcu_tasks_rude_kthread
    rcu_tasks_trace_kthread
    poll_mpt3sas0_statu
    ext4-rsv-conversion
    xfs-reclaim/sdf1
    xfs-blockgc/sdf1
    xfs-inodegc/sdf1
    audit_send_reply
    ecryptfs-kthread
    vfio-irqfd-cleanup
    jbd2/nvme0n1p2-8

I do wonder if that could break some user space that assumes these names have maximum length ..

But LGTM

Signature · 3 lines
Thanks,

David / dhildenb
#3 lore raw
Quoted 58 lines · David Hildenbrand
Quoted 44 lines · Yafang Shao

When I was implementing a new per-cpu kthread cfs_migration, I found the comm of it "cfs_migration/%u" is truncated due to the limitation of TASK_COMM_LEN. For example, the comm of the percpu thread on CPU10~19 are all with the same name "cfs_migration/1", which will confuse the user. This issue is not critical, because we can get the corresponding CPU from the task's Cpus_allowed. But for kthreads correspoinding to other hardware devices, it is not easy to get the detailed device info from task comm, for example,

    jbd2/nvme0n1p2-
    xfs-reclaim/sdf

Currently there are so many truncated kthreads:

    rcu_tasks_kthre
    rcu_tasks_rude_
    rcu_tasks_trace
    poll_mpt3sas0_s
    ext4-rsv-conver
    xfs-reclaim/sd{a, b, c, ...}
    xfs-blockgc/sd{a, b, c, ...}
    xfs-inodegc/sd{a, b, c, ...}
    audit_send_repl
    ecryptfs-kthrea
    vfio-irqfd-clea
    jbd2/nvme0n1p2-
    ...

We can shorten these names to work around this problem, but it may be not applied to all of the truncated kthreads. Take 'jbd2/nvme0n1p2-' for example, it is a nice name, and it is not a good idea to shorten it.

One possible way to fix this issue is extending the task comm size, but as task->comm is used in lots of places, that may cause some potential buffer overflows. Another more conservative approach is introducing a new pointer to store kthread's full name if it is truncated, which won't introduce too much overhead as it is in the non-critical path. Finally we make a dicision to use the second approach. See also the discussions in this thread:
https://lore.kernel.org/lkml/[email protected]/

After this change, the full name of these truncated kthreads will be displayed via /proc/[pid]/comm:

Quoted 12 lines
    rcu_tasks_kthread
    rcu_tasks_rude_kthread
    rcu_tasks_trace_kthread
    poll_mpt3sas0_statu
    ext4-rsv-conversion
    xfs-reclaim/sdf1
    xfs-blockgc/sdf1
    xfs-inodegc/sdf1
    audit_send_reply
    ecryptfs-kthread
    vfio-irqfd-cleanup
    jbd2/nvme0n1p2-8
Quoted 2 lines

I do wonder if that could break some user space that assumes these names have maximum length ..

There is high chance that we will be on the safe side. Workqueue kthreads already provided longer names. They are even dynamic because the currently handled workqueue name is part of the name, see wq_worker_comm().

Best Regards,
Petr

#4 lore raw
Quoted 59 lines · Petr Mladek
Quoted 58 lines · David Hildenbrand
Quoted 56 lines · Yafang Shao

When I was implementing a new per-cpu kthread cfs_migration, I found the comm of it "cfs_migration/%u" is truncated due to the limitation of TASK_COMM_LEN. For example, the comm of the percpu thread on CPU10~19 are all with the same name "cfs_migration/1", which will confuse the user. This issue is not critical, because we can get the corresponding CPU from the task's Cpus_allowed. But for kthreads correspoinding to other hardware devices, it is not easy to get the detailed device info from task comm, for example,

    jbd2/nvme0n1p2-
    xfs-reclaim/sdf

Currently there are so many truncated kthreads:

    rcu_tasks_kthre
    rcu_tasks_rude_
    rcu_tasks_trace
    poll_mpt3sas0_s
    ext4-rsv-conver
    xfs-reclaim/sd{a, b, c, ...}
    xfs-blockgc/sd{a, b, c, ...}
    xfs-inodegc/sd{a, b, c, ...}
    audit_send_repl
    ecryptfs-kthrea
    vfio-irqfd-clea
    jbd2/nvme0n1p2-
    ...

We can shorten these names to work around this problem, but it may be not applied to all of the truncated kthreads. Take 'jbd2/nvme0n1p2-' for example, it is a nice name, and it is not a good idea to shorten it.

One possible way to fix this issue is extending the task comm size, but as task->comm is used in lots of places, that may cause some potential buffer overflows. Another more conservative approach is introducing a new pointer to store kthread's full name if it is truncated, which won't introduce too much overhead as it is in the non-critical path. Finally we make a dicision to use the second approach. See also the discussions in this thread:
https://lore.kernel.org/lkml/[email protected]/

After this change, the full name of these truncated kthreads will be displayed via /proc/[pid]/comm:

    rcu_tasks_kthread
    rcu_tasks_rude_kthread
    rcu_tasks_trace_kthread
    poll_mpt3sas0_statu
    ext4-rsv-conversion
    xfs-reclaim/sdf1
    xfs-blockgc/sdf1
    xfs-inodegc/sdf1
    audit_send_reply
    ecryptfs-kthread
    vfio-irqfd-cleanup
    jbd2/nvme0n1p2-8
Quoted 7 lines
Quoted 2 lines

I do wonder if that could break some user space that assumes these names have maximum length ..

There is high chance that we will be on the safe side. Workqueue kthreads already provided longer names. They are even dynamic because the currently handled workqueue name is part of the name, see wq_worker_comm().

Great, thanks!

Signature · 3 lines
Thanks,

David / dhildenb