Skip to contentSkip to conversation

bpf: fix a couple of issues with syscall program

PATCH Superseded Yonghong Song 4 participants 8 messages Aug 9, 2021 – Aug 10, 2021

Subsystems (2)

v3 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

v3 cover bpf: fix a couple of issues with syscall program

syzbot ([1]) reported a rcu warning for syscall program when bpf_get_current_cgroup_id() is called in the program. The bpf_get_current_cgroup_id() helper expects cgroup related mutex or rcu_read_lock(). The sleepable program cannot be protected with rcu read_lock, hence the warning.

To fix the problem, Patch 1 added rcu_read_lock() in affected helpers, and Patch 2 added missing bpf_read_lock_trace() for syscall type of bpf programs.

[1] https://lore.kernel.org/bpf/[email protected]/

Changelog:
v2 -> v3:
- use rcu_read_lock() protection for bpf_get_current_[ancestor_]cgroup_id() helper.
v1 -> v2:
- use bpf_read_lock_trace() instead of bpf_read_lock() for bpf_prog_run_pin_on_cpu().
- disallow bpf_get_current_[ancestor_]cgroup_id() helper.

Yonghong Song (2):
bpf: add rcu read_lock in bpf_get_current_[ancestor_]cgroup_id()
helpers
bpf: add missing bpf_read_[un]lock_trace() for syscall program

 kernel/bpf/helpers.c | 12 ++++++++++--
 net/bpf/test_run.c   |  4 ++++
 2 files changed, 14 insertions(+), 2 deletions(-)
Signature · 1 line
2.30.2
#1 lore raw

PATCH v3 1/2 bpf: add rcu read_lock in bpf_get_current_[ancestor_]cgroup_id() helpers

Currently, if bpf_get_current_cgroup_id() or bpf_get_current_ancestor_cgroup_id() helper is called with sleepable programs e.g., sleepable fentry/fmod_ret/fexit/lsm programs, a rcu warning may appear. For example, if I added the following hack to test_progs/test_lsm sleepable fentry program test_sys_setdomainname:

  --- a/tools/testing/selftests/bpf/progs/lsm.c
  +++ b/tools/testing/selftests/bpf/progs/lsm.c
  @@ -168,6 +168,10 @@ int BPF_PROG(test_sys_setdomainname, struct pt_regs *regs)
          int buf = 0;
          long ret;

+ __u64 cg_id = bpf_get_current_cgroup_id();
+ if (cg_id == 1000)
+ copy_test++;
+

          ret = bpf_copy_from_user(&buf, sizeof(buf), ptr);
          if (len == -2 && ret == 0 && buf == 1234)
                  copy_test++;

I will hit the following rcu warning:

include/linux/cgroup.h:481 suspicious rcu_dereference_check() usage! other info that might help us debug this:

    rcu_scheduler_active = 2, debug_locks = 1
    1 lock held by test_progs/260:
      #0: ffffffffa5173360 (rcu_read_lock_trace){....}-{0:0}, at: __bpf_prog_enter_sleepable+0x0/0xa0
    stack backtrace:
    CPU: 1 PID: 260 Comm: test_progs Tainted: G           O      5.14.0-rc2+ #176
    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
    Call Trace:
      dump_stack_lvl+0x56/0x7b
      bpf_get_current_cgroup_id+0x9c/0xb1
      bpf_prog_a29888d1c6706e09_test_sys_setdomainname+0x3e/0x89c
      bpf_trampoline_6442469132_0+0x2d/0x1000
      __x64_sys_setdomainname+0x5/0x110
      do_syscall_64+0x3a/0x80
      entry_SYSCALL_64_after_hwframe+0x44/0xae

I can get similar warning using bpf_get_current_ancestor_cgroup_id() helper. syzbot reported a similar issue in [1] for syscall program. Helper bpf_get_current_cgroup_id() or bpf_get_current_ancestor_cgroup_id() has the following callchain:
task_dfl_cgroup
task_css_set
task_css_set_check
and we have

   #define task_css_set_check(task, __c)                                   \
           rcu_dereference_check((task)->cgroups,                          \
                   lockdep_is_held(&cgroup_mutex) ||                       \
                   lockdep_is_held(&css_set_lock) ||                       \
                   ((task)->flags & PF_EXITING) || (__c))

Since cgroup_mutex/css_set_lock is not held and the task is not existing and rcu read_lock is not held, a warning will be issued. Note that bpf sleepable program is protected by rcu_read_lock_trace().

The above sleepable bpf programs are already protected
by migrate_disable(). Adding rcu_read_lock() in these
two helpers will silence the above warning.
I marked the patch fixing 95b861a7935b
("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing") which added bpf_get_current_ancestor_cgroup_id() to tracing programs in 5.14. I think backporting 5.14 is probably good enough as sleepable progrems are not widely used.

This patch should fix [1] as well since syscall program is a sleepable program protected with migrate_disable().

[1] https://lore.kernel.org/bpf/[email protected]/

Reported-by [email protected]
Fixes 95b861a7935b ("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing")
---
 kernel/bpf/helpers.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)
kernel/bpf/helpers.c +10 -2
index 62cf00383910..4567d2841133 100644
@@ -353,7 +353,11 @@ const struct bpf_func_proto bpf_jiffies64_proto = {
 #ifdef CONFIG_CGROUPS
 BPF_CALL_0(bpf_get_current_cgroup_id)
 {
-	struct cgroup *cgrp = task_dfl_cgroup(current);
+	struct cgroup *cgrp;
+
+	rcu_read_lock();
+	cgrp = task_dfl_cgroup(current);
+	rcu_read_unlock();
 
 	return cgroup_id(cgrp);
 }
@@ -366,9 +370,13 @@ const struct bpf_func_proto bpf_get_current_cgroup_id_proto = {
 
 BPF_CALL_1(bpf_get_current_ancestor_cgroup_id, int, ancestor_level)
 {
-	struct cgroup *cgrp = task_dfl_cgroup(current);
+	struct cgroup *cgrp;
 	struct cgroup *ancestor;
 
+	rcu_read_lock();
+	cgrp = task_dfl_cgroup(current);
+	rcu_read_unlock();
+
 	ancestor = cgroup_ancestor(cgrp, ancestor_level);
 	if (!ancestor)
 		return 0;
Signature · 1 line
2.30.2
#6 lore raw

PATCH v3 2/2 bpf: add missing bpf_read_[un]lock_trace() for syscall program

In reply to Yonghong Song

Commit 79a7f8bdb159d ("bpf: Introduce bpf_sys_bpf() helper and program type.") added support for syscall program, which is a sleepable program. But the program run missed bpf_read_lock_trace()/bpf_read_unlock_trace(), which is needed to ensure proper rcu callback invocations.
This patch added bpf_read_[un]lock_trace() properly.

Fixes 79a7f8bdb159d ("bpf: Introduce bpf_sys_bpf() helper and program type.")
---
 net/bpf/test_run.c | 4 ++++
 1 file changed, 4 insertions(+)
net/bpf/test_run.c +4 -0
index 1cc75c811e24..caa16bf30fb5 100644
@@ -7,6 +7,7 @@
 #include <linux/vmalloc.h>
 #include <linux/etherdevice.h>
 #include <linux/filter.h>
+#include <linux/rcupdate_trace.h>
 #include <linux/sched/signal.h>
 #include <net/bpf_sk_storage.h>
 #include <net/sock.h>
@@ -951,7 +952,10 @@ int bpf_prog_test_run_syscall(struct bpf_prog *prog,
 			goto out;
 		}
 	}
+
+	rcu_read_lock_trace();
 	retval = bpf_prog_run_pin_on_cpu(prog, ctx);
+	rcu_read_unlock_trace();
 
 	if (copy_to_user(&uattr->test.retval, &retval, sizeof(u32))) {
 		err = -EFAULT;
Signature · 1 line
2.30.2
Acked-by on v3 1/2
#2 lore raw

Re: [PATCH bpf v3 1/2] bpf: add rcu read_lock in bpf_get_current_[ancestor_]cgroup_id() helpers

In reply to Yonghong Song
Quoted 72 lines · Yonghong Song

Currently, if bpf_get_current_cgroup_id() or bpf_get_current_ancestor_cgroup_id() helper is called with sleepable programs e.g., sleepable fentry/fmod_ret/fexit/lsm programs, a rcu warning may appear. For example, if I added the following hack to test_progs/test_lsm sleepable fentry program test_sys_setdomainname:

  --- a/tools/testing/selftests/bpf/progs/lsm.c
  +++ b/tools/testing/selftests/bpf/progs/lsm.c
  @@ -168,6 +168,10 @@ int BPF_PROG(test_sys_setdomainname, struct pt_regs *regs)
          int buf = 0;
          long ret;

+ __u64 cg_id = bpf_get_current_cgroup_id();
+ if (cg_id == 1000)
+ copy_test++;
+

          ret = bpf_copy_from_user(&buf, sizeof(buf), ptr);
          if (len == -2 && ret == 0 && buf == 1234)
                  copy_test++;

I will hit the following rcu warning:

include/linux/cgroup.h:481 suspicious rcu_dereference_check() usage! other info that might help us debug this:

    rcu_scheduler_active = 2, debug_locks = 1
    1 lock held by test_progs/260:
      #0: ffffffffa5173360 (rcu_read_lock_trace){....}-{0:0}, at: __bpf_prog_enter_sleepable+0x0/0xa0
    stack backtrace:
    CPU: 1 PID: 260 Comm: test_progs Tainted: G           O      5.14.0-rc2+ #176
    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
    Call Trace:
      dump_stack_lvl+0x56/0x7b
      bpf_get_current_cgroup_id+0x9c/0xb1
      bpf_prog_a29888d1c6706e09_test_sys_setdomainname+0x3e/0x89c
      bpf_trampoline_6442469132_0+0x2d/0x1000
      __x64_sys_setdomainname+0x5/0x110
      do_syscall_64+0x3a/0x80
      entry_SYSCALL_64_after_hwframe+0x44/0xae

I can get similar warning using bpf_get_current_ancestor_cgroup_id() helper. syzbot reported a similar issue in [1] for syscall program. Helper bpf_get_current_cgroup_id() or bpf_get_current_ancestor_cgroup_id() has the following callchain:
task_dfl_cgroup
task_css_set
task_css_set_check
and we have

   #define task_css_set_check(task, __c)                                   \
           rcu_dereference_check((task)->cgroups,                          \
                   lockdep_is_held(&cgroup_mutex) ||                       \
                   lockdep_is_held(&css_set_lock) ||                       \
                   ((task)->flags & PF_EXITING) || (__c))

Since cgroup_mutex/css_set_lock is not held and the task is not existing and rcu read_lock is not held, a warning will be issued. Note that bpf sleepable program is protected by rcu_read_lock_trace().

The above sleepable bpf programs are already protected
by migrate_disable(). Adding rcu_read_lock() in these
two helpers will silence the above warning.
I marked the patch fixing 95b861a7935b
("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing") which added bpf_get_current_ancestor_cgroup_id() to tracing programs in 5.14. I think backporting 5.14 is probably good enough as sleepable progrems are not widely used.

This patch should fix [1] as well since syscall program is a sleepable program protected with migrate_disable().

Quoted 6 lines

[1] https://lore.kernel.org/bpf/[email protected]/

Reported-by [email protected]
Fixes 95b861a7935b ("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing")

---

LGTM, thanks!

Quoted 38 lines
 kernel/bpf/helpers.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)
kernel/bpf/helpers.c +10 -3
index 62cf00383910..4567d2841133 100644
@@ -353,7 +353,11 @@ const struct bpf_func_proto bpf_jiffies64_proto = {
 #ifdef CONFIG_CGROUPS
 BPF_CALL_0(bpf_get_current_cgroup_id)
 {
-       struct cgroup *cgrp = task_dfl_cgroup(current);
+       struct cgroup *cgrp;
+
+       rcu_read_lock();
+       cgrp = task_dfl_cgroup(current);
+       rcu_read_unlock();
 
        return cgroup_id(cgrp);
 }
@@ -366,9 +370,13 @@ const struct bpf_func_proto bpf_get_current_cgroup_id_proto = {
 
 BPF_CALL_1(bpf_get_current_ancestor_cgroup_id, int, ancestor_level)
 {
-       struct cgroup *cgrp = task_dfl_cgroup(current);
+       struct cgroup *cgrp;
        struct cgroup *ancestor;
 
+       rcu_read_lock();
+       cgrp = task_dfl_cgroup(current);
+       rcu_read_unlock();
+
        ancestor = cgroup_ancestor(cgrp, ancestor_level);
        if (!ancestor)
                return 0;
--

2.30.2

maintainer
#3 lore raw

Re: [PATCH bpf v3 1/2] bpf: add rcu read_lock in bpf_get_current_[ancestor_]cgroup_id() helpers

In reply to Yonghong Song

[ +Paul ]

On 8/10/21 1:51 AM, Yonghong Song wrote:
[...]

Quoted 64 lines

I will hit the following rcu warning:

   include/linux/cgroup.h:481 suspicious rcu_dereference_check() usage!
   other info that might help us debug this:
     rcu_scheduler_active = 2, debug_locks = 1
     1 lock held by test_progs/260:
       #0: ffffffffa5173360 (rcu_read_lock_trace){....}-{0:0}, at: __bpf_prog_enter_sleepable+0x0/0xa0
     stack backtrace:
     CPU: 1 PID: 260 Comm: test_progs Tainted: G           O      5.14.0-rc2+ #176
     Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
     Call Trace:
       dump_stack_lvl+0x56/0x7b
       bpf_get_current_cgroup_id+0x9c/0xb1
       bpf_prog_a29888d1c6706e09_test_sys_setdomainname+0x3e/0x89c
       bpf_trampoline_6442469132_0+0x2d/0x1000
       __x64_sys_setdomainname+0x5/0x110
       do_syscall_64+0x3a/0x80
       entry_SYSCALL_64_after_hwframe+0x44/0xae

I can get similar warning using bpf_get_current_ancestor_cgroup_id() helper. syzbot reported a similar issue in [1] for syscall program. Helper bpf_get_current_cgroup_id() or bpf_get_current_ancestor_cgroup_id() has the following callchain:
task_dfl_cgroup
task_css_set
task_css_set_check
and we have

    #define task_css_set_check(task, __c)                                   \
            rcu_dereference_check((task)->cgroups,                          \
                    lockdep_is_held(&cgroup_mutex) ||                       \
                    lockdep_is_held(&css_set_lock) ||                       \
                    ((task)->flags & PF_EXITING) || (__c))

Since cgroup_mutex/css_set_lock is not held and the task is not existing and rcu read_lock is not held, a warning will be issued. Note that bpf sleepable program is protected by rcu_read_lock_trace().

The above sleepable bpf programs are already protected
by migrate_disable(). Adding rcu_read_lock() in these
two helpers will silence the above warning.
I marked the patch fixing 95b861a7935b
("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing") which added bpf_get_current_ancestor_cgroup_id() to tracing programs in 5.14. I think backporting 5.14 is probably good enough as sleepable progrems are not widely used.

This patch should fix [1] as well since syscall program is a sleepable program protected with migrate_disable().

[1] https://lore.kernel.org/bpf/[email protected]/

Reported-by [email protected]
Fixes 95b861a7935b ("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing")
---
  kernel/bpf/helpers.c | 12 ++++++++++--
  1 file changed, 10 insertions(+), 2 deletions(-)
kernel/bpf/helpers.c +0 -0
index 62cf00383910..4567d2841133 100644
@@ -353,7 +353,11 @@ const struct bpf_func_proto bpf_jiffies64_proto = {
  #ifdef CONFIG_CGROUPS

kernel/bpf/helpers.c const struct bpf_func_proto bpf_jiffies64_proto = { · line 362

Quoted 10 lines
kernel/bpf/helpers.c +5 -1
  BPF_CALL_0(bpf_get_current_cgroup_id)
  {
-	struct cgroup *cgrp = task_dfl_cgroup(current);
+	struct cgroup *cgrp;
+
+	rcu_read_lock();
+	cgrp = task_dfl_cgroup(current);
+	rcu_read_unlock();
 
  	return cgroup_id(cgrp);

I'm a bit confused, if cgroup object relies rcu_read_lock() and not rcu_read_lock_trace() context, then the above is racy given you access the memory via cgroup_id() outside of it, same below. If rcu_read_lock_trace() is enough and the above is really just to silence the 'suspicious rcu_dereference_check() usage' splat, then the rcu_dereference_check() from task_css_set_check() should be extended to check for _trace() flavor instead [meaning, as a cleaner workaround], which one is it?

Quoted 7 lines
kernel/bpf/helpers.c +0 -1
  }
@@ -366,9 +370,13 @@ const struct bpf_func_proto bpf_get_current_cgroup_id_proto = {
 
  BPF_CALL_1(bpf_get_current_ancestor_cgroup_id, int, ancestor_level)
  {
-	struct cgroup *cgrp = task_dfl_cgroup(current);

kernel/bpf/helpers.c const struct bpf_func_proto bpf_get_current_cgroup_id_proto = { · line 382

Quoted 10 lines
kernel/bpf/helpers.c +5 -0
+	struct cgroup *cgrp;
  	struct cgroup *ancestor;
 
+	rcu_read_lock();
+	cgrp = task_dfl_cgroup(current);
+	rcu_read_unlock();
+
  	ancestor = cgroup_ancestor(cgrp, ancestor_level);
  	if (!ancestor)
  		return 0;

Thanks,
Daniel

maintainer
#7 lore raw

Re: [PATCH bpf v3 2/2] bpf: add missing bpf_read_[un]lock_trace() for syscall program

In reply to Yonghong Song
Quoted 9 lines · Yonghong Song

Commit 79a7f8bdb159d ("bpf: Introduce bpf_sys_bpf() helper and program type.") added support for syscall program, which is a sleepable program. But the program run missed bpf_read_lock_trace()/bpf_read_unlock_trace(), which is needed to ensure proper rcu callback invocations.
This patch added bpf_read_[un]lock_trace() properly.

Fixes 79a7f8bdb159d ("bpf: Introduce bpf_sys_bpf() helper and program type.")

(Took this one in already while we clarify 1/2, thanks!)

#4 lore raw

Re: [PATCH bpf v3 1/2] bpf: add rcu read_lock in bpf_get_current_[ancestor_]cgroup_id() helpers

In reply to Daniel Borkmann
Quoted 76 lines · Daniel Borkmann

[ +Paul ]

On 8/10/21 1:51 AM, Yonghong Song wrote:
[...]

Quoted 72 lines

I will hit the following rcu warning:

   include/linux/cgroup.h:481 suspicious rcu_dereference_check() usage!
   other info that might help us debug this:
     rcu_scheduler_active = 2, debug_locks = 1
     1 lock held by test_progs/260:
       #0: ffffffffa5173360 (rcu_read_lock_trace){....}-{0:0}, at: __bpf_prog_enter_sleepable+0x0/0xa0
     stack backtrace:
     CPU: 1 PID: 260 Comm: test_progs Tainted: G           O      5.14.0-rc2+ #176
     Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
     Call Trace:
       dump_stack_lvl+0x56/0x7b
       bpf_get_current_cgroup_id+0x9c/0xb1
       bpf_prog_a29888d1c6706e09_test_sys_setdomainname+0x3e/0x89c
       bpf_trampoline_6442469132_0+0x2d/0x1000
       __x64_sys_setdomainname+0x5/0x110
       do_syscall_64+0x3a/0x80
       entry_SYSCALL_64_after_hwframe+0x44/0xae

I can get similar warning using bpf_get_current_ancestor_cgroup_id() helper. syzbot reported a similar issue in [1] for syscall program. Helper bpf_get_current_cgroup_id() or bpf_get_current_ancestor_cgroup_id() has the following callchain:
task_dfl_cgroup
task_css_set
task_css_set_check
and we have

    #define task_css_set_check(task, __c)                                   \
            rcu_dereference_check((task)->cgroups,                          \
                    lockdep_is_held(&cgroup_mutex) ||                       \
                    lockdep_is_held(&css_set_lock) ||                       \
                    ((task)->flags & PF_EXITING) || (__c))

Since cgroup_mutex/css_set_lock is not held and the task is not existing and rcu read_lock is not held, a warning will be issued. Note that bpf sleepable program is protected by rcu_read_lock_trace().

The above sleepable bpf programs are already protected
by migrate_disable(). Adding rcu_read_lock() in these
two helpers will silence the above warning.
I marked the patch fixing 95b861a7935b
("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing") which added bpf_get_current_ancestor_cgroup_id() to tracing programs in 5.14. I think backporting 5.14 is probably good enough as sleepable progrems are not widely used.

This patch should fix [1] as well since syscall program is a sleepable program protected with migrate_disable().

[1] https://lore.kernel.org/bpf/[email protected]/

Reported-by [email protected]
Fixes 95b861a7935b ("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing")
---
  kernel/bpf/helpers.c | 12 ++++++++++--
  1 file changed, 10 insertions(+), 2 deletions(-)
kernel/bpf/helpers.c +5 -1
index 62cf00383910..4567d2841133 100644
@@ -353,7 +353,11 @@ const struct bpf_func_proto bpf_jiffies64_proto = {
  #ifdef CONFIG_CGROUPS
  BPF_CALL_0(bpf_get_current_cgroup_id)
  {
-	struct cgroup *cgrp = task_dfl_cgroup(current);
+	struct cgroup *cgrp;
+
+	rcu_read_lock();
+	cgrp = task_dfl_cgroup(current);
+	rcu_read_unlock();
Quoted 8 lines

kernel/bpf/helpers.c const struct bpf_func_proto bpf_jiffies64_proto = { · line 362

Quoted 1 line
kernel/bpf/helpers.c +0 -0
  	return cgroup_id(cgrp);

I'm a bit confused, if cgroup object relies rcu_read_lock() and not rcu_read_lock_trace() context, then the above is racy given you access the memory via cgroup_id() outside of it, same below. If rcu_read_lock_trace() is enough and the above is really just to silence the 'suspicious rcu_dereference_check() usage' splat, then the rcu_dereference_check() from task_css_set_check() should be extended to check for _trace() flavor instead [meaning, as a cleaner workaround], which one is it?

At first glance, something like this would work from an RCU perspective:

BPF_CALL_0(bpf_get_current_cgroup_id)

{
	struct cgroup *cgrp = task_dfl_cgroup(current);
	struct cgroup *cgrp;
	u64 ret;

	rcu_read_lock();
	cgrp = task_dfl_cgroup(current);
 	ret = cgroup_id(cgrp);
	rcu_read_unlock();

return ret;
}

If I am reading the code correctly, rcu_read_lock() is required to keep the cgroup from going away, and the rcu_read_lock_trace() is needed in addition in order to keep the BPF program from going away.

There might well be better ways to do this, but the above obeys the letter of the law. Or at least the law as I understand it. ;-)

Or does the fact that a cgroup contains a given task prevent that cgroup from going away? (I -think- that tasks can be removed from cgroups without the cooperation of those tasks, but then again there is much about cgroups that I do not know.)

Thanx, Paul

Quoted 18 lines
Quoted 5 lines
kernel/bpf/helpers.c +0 -1
  }
@@ -366,9 +370,13 @@ const struct bpf_func_proto bpf_get_current_cgroup_id_proto = {
  BPF_CALL_1(bpf_get_current_ancestor_cgroup_id, int, ancestor_level)
  {
-	struct cgroup *cgrp = task_dfl_cgroup(current);

kernel/bpf/helpers.c const struct bpf_func_proto bpf_get_current_cgroup_id_proto = { · line 381

Quoted 10 lines
kernel/bpf/helpers.c +5 -0
+	struct cgroup *cgrp;
  	struct cgroup *ancestor;
+	rcu_read_lock();
+	cgrp = task_dfl_cgroup(current);
+	rcu_read_unlock();
+
  	ancestor = cgroup_ancestor(cgrp, ancestor_level);
  	if (!ancestor)
  		return 0;
 

Thanks,
Daniel

#5 lore raw

Re: [PATCH bpf v3 1/2] bpf: add rcu read_lock in bpf_get_current_[ancestor_]cgroup_id() helpers

Add Tejun to provide more clarification on cgroup side.

Quoted 77 lines · Paul E. McKenney
Quoted 76 lines · Daniel Borkmann

[ +Paul ]

On 8/10/21 1:51 AM, Yonghong Song wrote:
[...]

Quoted 72 lines

I will hit the following rcu warning:

    include/linux/cgroup.h:481 suspicious rcu_dereference_check() usage!
    other info that might help us debug this:
      rcu_scheduler_active = 2, debug_locks = 1
      1 lock held by test_progs/260:
        #0: ffffffffa5173360 (rcu_read_lock_trace){....}-{0:0}, at: __bpf_prog_enter_sleepable+0x0/0xa0
      stack backtrace:
      CPU: 1 PID: 260 Comm: test_progs Tainted: G           O      5.14.0-rc2+ #176
      Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
      Call Trace:
        dump_stack_lvl+0x56/0x7b
        bpf_get_current_cgroup_id+0x9c/0xb1
        bpf_prog_a29888d1c6706e09_test_sys_setdomainname+0x3e/0x89c
        bpf_trampoline_6442469132_0+0x2d/0x1000
        __x64_sys_setdomainname+0x5/0x110
        do_syscall_64+0x3a/0x80
        entry_SYSCALL_64_after_hwframe+0x44/0xae

I can get similar warning using bpf_get_current_ancestor_cgroup_id() helper. syzbot reported a similar issue in [1] for syscall program. Helper bpf_get_current_cgroup_id() or bpf_get_current_ancestor_cgroup_id() has the following callchain:
task_dfl_cgroup
task_css_set
task_css_set_check
and we have

     #define task_css_set_check(task, __c)                                   \
             rcu_dereference_check((task)->cgroups,                          \
                     lockdep_is_held(&cgroup_mutex) ||                       \
                     lockdep_is_held(&css_set_lock) ||                       \
                     ((task)->flags & PF_EXITING) || (__c))

Since cgroup_mutex/css_set_lock is not held and the task is not existing and rcu read_lock is not held, a warning will be issued. Note that bpf sleepable program is protected by rcu_read_lock_trace().

The above sleepable bpf programs are already protected
by migrate_disable(). Adding rcu_read_lock() in these
two helpers will silence the above warning.
I marked the patch fixing 95b861a7935b
("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing") which added bpf_get_current_ancestor_cgroup_id() to tracing programs in 5.14. I think backporting 5.14 is probably good enough as sleepable progrems are not widely used.

This patch should fix [1] as well since syscall program is a sleepable program protected with migrate_disable().

   [1] https://lore.kernel.org/bpf/[email protected]/
Reported-by [email protected]
Fixes 95b861a7935b ("bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing")
---
   kernel/bpf/helpers.c | 12 ++++++++++--
   1 file changed, 10 insertions(+), 2 deletions(-)
kernel/bpf/helpers.c +5 -1
index 62cf00383910..4567d2841133 100644
@@ -353,7 +353,11 @@ const struct bpf_func_proto bpf_jiffies64_proto = {
   #ifdef CONFIG_CGROUPS
   BPF_CALL_0(bpf_get_current_cgroup_id)
   {
-	struct cgroup *cgrp = task_dfl_cgroup(current);
+	struct cgroup *cgrp;
+
+	rcu_read_lock();
+	cgrp = task_dfl_cgroup(current);
+	rcu_read_unlock();
Quoted 8 lines
Quoted 8 lines

kernel/bpf/helpers.c const struct bpf_func_proto bpf_jiffies64_proto = { · line 362

Quoted 1 line
kernel/bpf/helpers.c +0 -0
   	return cgroup_id(cgrp);

I'm a bit confused, if cgroup object relies rcu_read_lock() and not rcu_read_lock_trace() context, then the above is racy given you access the memory via cgroup_id() outside of it, same below. If rcu_read_lock_trace() is enough and the above is really just to silence the 'suspicious rcu_dereference_check() usage' splat, then the rcu_dereference_check() from task_css_set_check() should be extended to check for _trace() flavor instead [meaning, as a cleaner workaround], which one is it?

Thanks for pointing out. Definitely a bug on my side. rcu_read_lock() region should include cgroup_id(cgrp) as well.

Quoted 48 lines

At first glance, something like this would work from an RCU perspective:

BPF_CALL_0(bpf_get_current_cgroup_id)

{
	struct cgroup *cgrp = task_dfl_cgroup(current);
	struct cgroup *cgrp;
	u64 ret;

	rcu_read_lock();
	cgrp = task_dfl_cgroup(current);
  	ret = cgroup_id(cgrp);
	rcu_read_unlock();

return ret;
}

If I am reading the code correctly, rcu_read_lock() is required to keep the cgroup from going away, and the rcu_read_lock_trace() is needed in addition in order to keep the BPF program from going away.

There might well be better ways to do this, but the above obeys the letter of the law. Or at least the law as I understand it. ;-)

Or does the fact that a cgroup contains a given task prevent that cgroup from going away? (I -think- that tasks can be removed from cgroups without the cooperation of those tasks, but then again there is much about cgroups that I do not know.)

Thanx, Paul

Quoted 18 lines
Quoted 5 lines
kernel/bpf/helpers.c +0 -1
   }
@@ -366,9 +370,13 @@ const struct bpf_func_proto bpf_get_current_cgroup_id_proto = {
   BPF_CALL_1(bpf_get_current_ancestor_cgroup_id, int, ancestor_level)
   {
-	struct cgroup *cgrp = task_dfl_cgroup(current);

kernel/bpf/helpers.c const struct bpf_func_proto bpf_get_current_cgroup_id_proto = { · line 381

Quoted 10 lines
kernel/bpf/helpers.c +5 -0
+	struct cgroup *cgrp;
   	struct cgroup *ancestor;
+	rcu_read_lock();
+	cgrp = task_dfl_cgroup(current);
+	rcu_read_unlock();
+
   	ancestor = cgroup_ancestor(cgrp, ancestor_level);
   	if (!ancestor)
   		return 0;
 

Thanks,
Daniel