On 08/19, Daniel Borkmann wrote:
On 8/17/21 5:45 PM, Stanislav Fomichev wrote:
> Same as previous patch but for the keys. memdup_bpfptr is renamed
> to vmemdup_bpfptr (and converted to kvmalloc).
>
> Signed-off-by: Stanislav Fomichev <sdf@xxxxxxxxxx>
> ---
> include/linux/bpfptr.h | 12 ++++++++++--
> kernel/bpf/syscall.c | 34 +++++++++++++++++-----------------
> 2 files changed, 27 insertions(+), 19 deletions(-)
>
> diff --git a/include/linux/bpfptr.h b/include/linux/bpfptr.h
> index 5cdeab497cb3..84eeffb4316a 100644
> --- a/include/linux/bpfptr.h
> +++ b/include/linux/bpfptr.h
> @@ -62,9 +62,17 @@ static inline int copy_to_bpfptr_offset(bpfptr_t
dst, size_t offset,
> return copy_to_sockptr_offset((sockptr_t) dst, offset, src, size);
> }
> -static inline void *memdup_bpfptr(bpfptr_t src, size_t len)
> +static inline void *vmemdup_bpfptr(bpfptr_t src, size_t len)
nit: should we just name it kvmemdup_bpfptr() in that case?
Sounds good!
> {
> - return memdup_sockptr((sockptr_t) src, len);
> + void *p = kvmalloc(len, GFP_USER | __GFP_NOWARN);
> +
> + if (!p)
> + return ERR_PTR(-ENOMEM);
> + if (copy_from_sockptr(p, (sockptr_t) src, len)) {
Also, I think this one should rather use copy_from_bpfptr() here.
Ah, missed that one, thanks!
> + kvfree(p);
> + return ERR_PTR(-EFAULT);
> + }
> + return p;
> }
Rest lgtm, thanks!
