lkml.org 
[lkml]   [2022]   [Jan]   [31]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: [PATCH v7 bpf-next 8/9] bpf: introduce bpf_jit_binary_pack_[alloc|finalize|free]
From
Date
On 1/29/22 12:45 AM, Song Liu wrote:
[...]
> diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
> index 25e34caa9a95..ff0c51ef1cb7 100644
> --- a/kernel/bpf/core.c
> +++ b/kernel/bpf/core.c
> @@ -1031,6 +1031,109 @@ void bpf_jit_binary_free(struct bpf_binary_header *hdr)
> bpf_jit_uncharge_modmem(size);
> }
>
> +/* Allocate jit binary from bpf_prog_pack allocator.
> + * Since the allocated meory is RO+X, the JIT engine cannot write directly

nit: meory

> + * to the memory. To solve this problem, a RW buffer is also allocated at
> + * as the same time. The JIT engine should calculate offsets based on the
> + * RO memory address, but write JITed program to the RW buffer. Once the
> + * JIT engine finishes, it calls bpf_jit_binary_pack_finalize, which copies
> + * the JITed program to the RO memory.
> + */
> +struct bpf_binary_header *
> +bpf_jit_binary_pack_alloc(unsigned int proglen, u8 **image_ptr,
> + unsigned int alignment,
> + struct bpf_binary_header **rw_header,
> + u8 **rw_image,
> + bpf_jit_fill_hole_t bpf_fill_ill_insns)
> +{
> + struct bpf_binary_header *ro_header;
> + u32 size, hole, start;
> +
> + WARN_ON_ONCE(!is_power_of_2(alignment) ||
> + alignment > BPF_IMAGE_ALIGNMENT);
> +
> + /* add 16 bytes for a random section of illegal instructions */
> + size = round_up(proglen + sizeof(*ro_header) + 16, BPF_PROG_CHUNK_SIZE);
> +
> + if (bpf_jit_charge_modmem(size))
> + return NULL;
> + ro_header = bpf_prog_pack_alloc(size);
> + if (!ro_header) {
> + bpf_jit_uncharge_modmem(size);
> + return NULL;
> + }
> +
> + *rw_header = kvmalloc(size, GFP_KERNEL);
> + if (!*rw_header) {
> + bpf_prog_pack_free(ro_header);
> + bpf_jit_uncharge_modmem(size);
> + return NULL;
> + }
> +
> + /* Fill space with illegal/arch-dep instructions. */
> + bpf_fill_ill_insns(*rw_header, size);
> + (*rw_header)->size = size;
> +
> + hole = min_t(unsigned int, size - (proglen + sizeof(*ro_header)),
> + BPF_PROG_CHUNK_SIZE - sizeof(*ro_header));
> + start = (get_random_int() % hole) & ~(alignment - 1);
> +
> + *image_ptr = &ro_header->image[start];
> + *rw_image = &(*rw_header)->image[start];
> +
> + return ro_header;
> +}
> +
> +/* Copy JITed text from rw_header to its final location, the ro_header. */
> +int bpf_jit_binary_pack_finalize(struct bpf_prog *prog,
> + struct bpf_binary_header *ro_header,
> + struct bpf_binary_header *rw_header)
> +{
> + void *ptr;
> +
> + ptr = bpf_arch_text_copy(ro_header, rw_header, rw_header->size);

Does this need to be wrapped with a text_mutex lock/unlock pair given
text_poke_copy() internally relies on __text_poke() ?

> + kvfree(rw_header);
> +
> + if (IS_ERR(ptr)) {
> + bpf_prog_pack_free(ro_header);
> + return PTR_ERR(ptr);
> + }
> + prog->aux->use_bpf_prog_pack = true;
> + return 0;
> +}
> +
[...]

\
 
 \ /
  Last update: 2022-02-01 01:22    [W:0.116 / U:0.040 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site