Currently passing alignment greater than 4 to bpf_jit_binary_alloc does
not work: in such cases it aligns only to 4 bytes.
However, this is required on s390, where in order to load a constant
from memory in a large (>512k) BPF program, one must use lgrl
instruction, whose memory operand must be aligned on an 8-byte boundary.
This patch makes it possible to request an arbitrary power-of-2
alignment from bpf_jit_binary_alloc by allocating extra padding bytes
and aligning the resulting pointer rather than the start offset.
An alternative would be to simply increase the alignment of
bpf_binary_header.image to 8, but this would increase the risk of
wasting a page on arches that don't need it, and would also be
insufficient in case someone needs e.g. 16-byte alignment in the
future.
Signed-off-by: Ilya Leoshkevich <iii@xxxxxxxxxxxxx>
---
include/linux/filter.h | 6 ++++--
kernel/bpf/core.c | 22 +++++++++++++++++-----
2 files changed, 21 insertions(+), 7 deletions(-)
diff --git a/include/linux/filter.h b/include/linux/filter.h
index 7a6f8f6f1da4..351a31eec24b 100644
--- a/include/linux/filter.h
+++ b/include/linux/filter.h
@@ -515,10 +515,12 @@ struct sock_fprog_kern {
struct sock_filter *filter;
};
+/* Some arches need word alignment for their instructions */
+#define BPF_IMAGE_ALIGNMENT 4
+
struct bpf_binary_header {
u32 pages;
- /* Some arches need word alignment for their instructions */
- u8 image[] __aligned(4);
+ u8 image[] __aligned(BPF_IMAGE_ALIGNMENT);
};
struct bpf_prog {
diff --git a/kernel/bpf/core.c b/kernel/bpf/core.c
index c1fde0303280..75dd3a43ada0 100644
--- a/kernel/bpf/core.c
+++ b/kernel/bpf/core.c
@@ -31,6 +31,8 @@
#include <linux/rcupdate.h>
#include <linux/perf_event.h>
#include <linux/extable.h>
+#include <linux/kernel.h>
+#include <linux/log2.h>
#include <asm/unaligned.h>
/* Registers */
@@ -812,14 +814,20 @@ bpf_jit_binary_alloc(unsigned int proglen, u8 **image_ptr,
unsigned int alignment,
bpf_jit_fill_hole_t bpf_fill_ill_insns)
{
+ u32 size, hole, start, pages, padding;
struct bpf_binary_header *hdr;
- u32 size, hole, start, pages;
+
+ WARN_ON_ONCE(!is_power_of_2(alignment));
+ if (alignment <= BPF_IMAGE_ALIGNMENT)
+ padding = 0;
+ else
+ padding = alignment - BPF_IMAGE_ALIGNMENT + 1;
/* Most of BPF filters are really small, but if some of them
* fill a page, allow at least 128 extra bytes to insert a
* random section of illegal instructions.
*/
- size = round_up(proglen + sizeof(*hdr) + 128, PAGE_SIZE);
+ size = round_up(proglen + sizeof(*hdr) + padding + 128, PAGE_SIZE);
pages = size / PAGE_SIZE;
if (bpf_jit_charge_modmem(pages))
@@ -834,12 +842,16 @@ bpf_jit_binary_alloc(unsigned int proglen, u8 **image_ptr,
bpf_fill_ill_insns(hdr, size);
hdr->pages = pages;
- hole = min_t(unsigned int, size - (proglen + sizeof(*hdr)),
+ hole = min_t(unsigned int,
+ size - (proglen + sizeof(*hdr) + padding),
PAGE_SIZE - sizeof(*hdr));
- start = (get_random_int() % hole) & ~(alignment - 1);
+ start = get_random_int() % hole;
/* Leave a random number of instructions before BPF code. */
- *image_ptr = &hdr->image[start];
+ if (alignment <= BPF_IMAGE_ALIGNMENT)
+ *image_ptr = &hdr->image[start & ~(alignment - 1)];
+ else
+ *image_ptr = PTR_ALIGN(&hdr->image[start], alignment);
return hdr;
}
--
2.23.0
