drbd: reject data replies with an out-of-range payload size
Subsystems (2)
PATCH 1/1 drbd: reject data replies with an out-of-range payload size
recv_dless_read() receives a P_DATA_REPLY from a peer into the bio of an outstanding read request. The peer-supplied payload length reaches it as the signed int data_size, and two peer-controlled inputs can make it negative. With a negotiated data-integrity-alg the digest length is subtracted first, so a reply whose payload is smaller than the digest underflows data_size. With no integrity algorithm (the default) data_size is assigned from the unsigned h95/h100 wire length and drbdd() never bounds it for a payload-carrying command, so a length above INT_MAX casts it negative; this path needs no non-default feature. The bio receive loop then computes expect = min_t(int, data_size, bv_len), which is negative, and drbd_recv_all_warn(mapped, expect) receives with a size_t of SIZE_MAX into the first mapped page.
The sibling receive path read_in_block() is not affected: it uses an unsigned size and rejects it against DRBD_MAX_BIO_SIZE before receiving. Reject a data reply whose size is negative after the optional digest subtraction, covering both triggers.
Impact: a malicious or man-in-the-middle DRBD peer copies attacker-chosen bytes past a bio page in the receiver, corrupting kernel memory. A node that reads from its peer (a diskless node, or read-balancing to the peer) is exposed in the default configuration; data-integrity-alg is not required.
---
Notes:
- Reproduced on mainline f5459048c38a in a virtme-ng KASAN guest driving the real receiver over a loopback DRBD 8.4 connection with a man-in-the-middle proxy. Both triggers were confirmed before/after:
(1) with data-integrity-alg=sha256, a P_DATA_REPLY one byte shorter than the digest; (2) with NO integrity algorithm (default), a P_DATA_REPLY whose h100 length is rewritten above INT_MAX. Stock panics with "KASAN: use-after-free in _copy_to_iter" from receive_DataReply (data_size seen as -1 and -1073741848 respectively); the patched build rejects both with -EIO and no KASAN. A benign read control (unmodified replies up to 128 KiB) passes on both stock and patched. The harness is available on request.
- The sibling receive path read_in_block() is not affected: it uses an unsigned size and rejects it against DRBD_MAX_BIO_SIZE before receiving.
- The pending "drbd: rework receiver for DRBD 9 transport and multi-peer protocol" series keeps the same unguarded subtraction and signed size in recv_dless_read(), so this check needs to be carried there too.
drivers/block/drbd/drbd_receiver.c | 5 +++++ 1 file changed, 5 insertions(+)
drivers/block/drbd/drbd_receiver.c +5 -0
@@ -1810,6 +1810,11 @@ static int recv_dless_read(struct drbd_peer_device *peer_device, struct drbd_req data_size -= digest_size; } +if (data_size <0 ) { +drbd_err (peer_device,"Invalid data reply size\n" ); +return -EIO ; + } +/* optimistically update recv_cnt. if receiving fails below, * we disconnect anyways, and counters will be reset. */ peer_device->device->recv_cnt += data_size>>9 ;
Signature · 1 line
2.53.0
Quoted 24 lines · Michael Bommarito
recv_dless_read()receives aP_DATA_REPLYfrom a peer into the bio of an outstanding read request. The peer-supplied payload length reaches it as the signed intdata_size, and two peer-controlled inputs can make it negative. With a negotiated data-integrity-alg the digest length is subtracted first, so a reply whose payload is smaller than the digest underflowsdata_size. With no integrity algorithm (the default)data_sizeis assigned from the unsigned h95/h100 wire length anddrbdd()never bounds it for a payload-carrying command, so a length aboveINT_MAXcasts it negative; this path needs no non-default feature. The bio receive loop then computes expect =min_t(int, data_size, bv_len), which is negative, anddrbd_recv_all_warn(mapped, expect)receives with asize_tofSIZE_MAXinto the first mapped page.The sibling receive path
read_in_block()is not affected: it uses an unsigned size and rejects it againstDRBD_MAX_BIO_SIZEbefore receiving. Reject a data reply whose size is negative after the optional digest subtraction, covering both triggers.Impact: a malicious or man-in-the-middle DRBD peer copies attacker-chosen bytes past a bio page in the receiver, corrupting kernel memory. A node that reads from its peer (a diskless node, or read-balancing to the peer) is exposed in the default configuration; data-integrity-alg is not required.
Quoted 5 lines
Fixes b411b3637fa7 ("The DRBD driver")Assisted-by Codex:gpt-5-5-xhighAssisted-by Claude:claude-opus-4-8Signed-off-by Michael Bommarito <[email protected]>
Looks correct, thanks.
Note that DRBD usually chooses to trusts its peers by design, but I agree that memory corruption is a bad enough consequence that we should put the guard there regardless.
Quoted 13 lines · Michael Bommarito
recv_dless_read()receives aP_DATA_REPLYfrom a peer into the bio of an outstanding read request. The peer-supplied payload length reaches it as the signed intdata_size, and two peer-controlled inputs can make it negative. With a negotiated data-integrity-alg the digest length is subtracted first, so a reply whose payload is smaller than the digest underflowsdata_size. With no integrity algorithm (the default)data_sizeis assigned from the unsigned h95/h100 wire length anddrbdd()never bounds it for a payload-carrying command, so a length aboveINT_MAXcasts it negative; this path needs no non-default feature. The bio receive loop then computes expect =min_t(int, data_size, bv_len), which is negative, anddrbd_recv_all_warn(mapped, expect)receives with asize_tofSIZE_MAXinto the first mapped page.
Quoted 1 line
[...]
Applied, thanks!
[1/1] drbd: reject data replies with an out-of-range payload size
commit: bd910a7660d280595ef94cb6d193951d855d330f
Best regards,
Signature · 1 line
Jens Axboe