[PATCH bpf-next] bpf: use kvmalloc in map_lookup_elem

Use kvmalloc/kvfree for temporary value when looking up a map.
kmalloc might not be sufficient for percpu maps where the value is big.

Can be reproduced with netcnt test on qemu with "-smp 255".

Signed-off-by: Stanislav Fomichev <redacted>
---
 kernel/bpf/syscall.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 9a2068e39d23..ae0b1c1c8ece 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c

@@ -1076,7 +1076,7 @@ static int map_lookup_elem(union bpf_attr *attr)
 	value_size = bpf_map_value_size(map);
 
 	err = -ENOMEM;
-	value = kmalloc(value_size, GFP_USER | __GFP_NOWARN);
+	value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN);
 	if (!value)
 		goto free_key;
 

@@ -1091,7 +1091,7 @@ static int map_lookup_elem(union bpf_attr *attr)
 	err = 0;
 
 free_value:
-	kfree(value);
+	kvfree(value);
 free_key:
 	kfree(key);
 err_put:

-- 
2.33.0.rc1.237.g0d66db33f3-goog

On 8/16/21 6:48 PM, Stanislav Fomichev wrote:

What about other cases like map_update_elem(), shouldn't they be adapted
similarly?

Thanks,
Daniel

On Mon, Aug 16, 2021 at 2:43 PM Daniel Borkmann [off-list ref] wrote:

On 8/16/21 6:48 PM, Stanislav Fomichev wrote:

quoted

Use kvmalloc/kvfree for temporary value when looking up a map.
kmalloc might not be sufficient for percpu maps where the value is big.

Can be reproduced with netcnt test on qemu with "-smp 255".

Signed-off-by: Stanislav Fomichev <redacted>
---
  kernel/bpf/syscall.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 9a2068e39d23..ae0b1c1c8ece 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c

@@ -1076,7 +1076,7 @@ static int map_lookup_elem(union bpf_attr *attr)
      value_size = bpf_map_value_size(map);

      err = -ENOMEM;
-     value = kmalloc(value_size, GFP_USER | __GFP_NOWARN);
+     value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN);
      if (!value)
              goto free_key;

What about other cases like map_update_elem(), shouldn't they be adapted
similarly?

And in the same vein (with keys potentially being big as well), should
we switch __bpf_copy_key() to use vmemdup_user() instead of
memdup_user()?

Thanks,
Daniel

On 08/16, Andrii Nakryiko wrote:

On Mon, Aug 16, 2021 at 2:43 PM Daniel Borkmann [off-list ref]  
wrote:

quoted

On 8/16/21 6:48 PM, Stanislav Fomichev wrote:

quoted

Use kvmalloc/kvfree for temporary value when looking up a map.
kmalloc might not be sufficient for percpu maps where the value is  

big.

quoted

quoted

Can be reproduced with netcnt test on qemu with "-smp 255".

Signed-off-by: Stanislav Fomichev <redacted>
---
  kernel/bpf/syscall.c | 4 ++--
  1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/bpf/syscall.c b/kernel/bpf/syscall.c
index 9a2068e39d23..ae0b1c1c8ece 100644
--- a/kernel/bpf/syscall.c
+++ b/kernel/bpf/syscall.c

@@ -1076,7 +1076,7 @@ static int map_lookup_elem(union bpf_attr *attr)
      value_size = bpf_map_value_size(map);

      err = -ENOMEM;
-     value = kmalloc(value_size, GFP_USER | __GFP_NOWARN);
+     value = kvmalloc(value_size, GFP_USER | __GFP_NOWARN);
      if (!value)
              goto free_key;

What about other cases like map_update_elem(), shouldn't they be adapted
similarly?

And in the same vein (with keys potentially being big as well), should
we switch __bpf_copy_key() to use vmemdup_user() instead of
memdup_user()?

Those are good questions :-)

I'm assuming that whatever is doing kmalloc on top of
bpf_map_value_size() should definitely use kvmalloc since
bpf_map_value_size() can blow up the value size. (will resend)

For __bpf_copy_key I'm less sure, but it might be a good idea as well.
Let me try to look at bit more into it, but feels like there shouldn't
be any downsides?