[PATCH v4] ceph: fix OOB read in ceph_osdc_list_watchers via uncapped outdata_len

[PATCH v4] ceph: fix OOB read in ceph_osdc_list_watchers via uncapped outdata_len

[Pavitra Jha]

The OSD reply header field op->payload_len is wire-controlled and is
copied directly into m->outdata_len[i] without any bounds check:

  m->outdata_len[i] = le32_to_cpu(op->payload_len);

This value propagates unchecked to req->r_ops[0].outdata_len and is
then used to set the decode boundary in ceph_osdc_list_watchers():

  void *const end = p + req->r_ops[0].outdata_len;

The actual data allocation is always exactly one page:
  ceph_alloc_page_vector(1, GFP_NOIO)
  ceph_osd_data_pages_init(..., PAGE_SIZE, ...)

The messenger caps the copy to PAGE_SIZE bytes, but the decode window
end is set from the uncapped wire value. A malicious OSD can send
outdata_len=0x10000, causing _safe decoder boundary checks to pass
while the physical reads cross the slab allocation boundary.
KASAN report (kernel 7.0.0-rc7, QEMU/x86_64, KASLR disabled):

  ==================================================================
  BUG: KASAN: slab-out-of-bounds in ceph_oob2_init+0x23d/0xff0 [ceph_oob2_poc]
  Read of size 4 at addr ffff88800a229f9e by task insmod/57

  CPU: 0 UID: 0 PID: 57 Comm: insmod Tainted: G           O        7.0.0-rc7-g9c2abf69da83-dirty #15 PREEMPT(lazy)
  Tainted: [O]=OOT_MODULE
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.17.0-debian-1.17.0-1 04/01/2014
  Call Trace:
   <TASK>
   dump_stack_lvl+0x4d/0x70
   print_report+0x170/0x4f3
   kasan_report+0xda/0x110
   ceph_oob2_init+0x23d/0xff0 [ceph_oob2_poc]
   do_one_initcall+0x9a/0x3a0
   do_init_module+0x27c/0x790
   load_module+0x4a9a/0x6350
   init_module_from_file+0x15c/0x180
   idempotent_init_module+0x21f/0x750
   __x64_sys_finit_module+0xba/0x120
   do_syscall_64+0xe2/0x570
   entry_SYSCALL_64_after_hwframe+0x77/0x7f

  Allocated by task 57:
   kasan_save_stack+0x30/0x50
   kasan_save_track+0x14/0x30
   __kasan_kmalloc+0x7f/0x90
   ceph_oob2_init+0x44/0xff0 [ceph_oob2_poc]
   do_one_initcall+0x9a/0x3a0
   do_init_module+0x27c/0x790
   load_module+0x4a9a/0x6350
   init_module_from_file+0x15c/0x180
   idempotent_init_module+0x21f/0x750
   __x64_sys_finit_module+0xba/0x120
   do_syscall_64+0xe2/0x570
   entry_SYSCALL_64_after_hwframe+0x77/0x7f

  The buggy address belongs to the object at ffff88800a229000
   which belongs to the cache kmalloc-4k of size 4096
  The buggy address is located 3998 bytes inside of
   allocated 4000-byte region [ffff88800a229000, ffff88800a229fa0)

  Memory state around the buggy address:
   ffff88800a229e80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   ffff88800a229f00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  >ffff88800a229f80: 00 00 00 00 fc fc fc fc fc fc fc fc fc fc fc fc
                                 ^
   ffff88800a22a000: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
   ffff88800a22a080: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
  ==================================================================

  val=0xccccaaaa (OOB garbage from KASAN redzone)

Fix by introducing buf_len to hold the allocation size, using it in
both ceph_osd_data_pages_init() and the min_t() decode boundary cap,
so the two are guaranteed to stay in sync if the buffer size changes.
buf_len is declared as u32 to match the type of outdata_len used in
the min_t() expression.

Attacker model: a malicious or compromised OSD in a multi-tenant
Ceph deployment can trigger this against any client issuing
CEPH_OSD_OP_LIST_WATCHERS without further privileges beyond OSD
session establishment.

Fixes: a4ed38d7a180 ("libceph: support for CEPH_OSD_OP_LIST_WATCHERS")
Cc: stable@vger.kernel.org
Reviewed-by: Viacheslav Dubeyko <slava@dubeyko.com>
Signed-off-by: Pavitra Jha <jhapavitra98@gmail.com>
---
v4: Rebase against current linux/master and ceph-testing/testing,
    confirmed clean git am -3 apply against both. No functional
    changes from Slava's reviewed v3.
v3: Change buf_len type from size_t to u32 to match outdata_len type
    in min_t(), per Viacheslav Dubeyko's review.
v2: Introduce buf_len variable instead of hardcoding PAGE_SIZE
    independently in ceph_osd_data_pages_init() and the min_t() cap,
    per Viacheslav Dubeyko's review.
---
 net/ceph/osd_client.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c
index 2ff00070c..62bec6ad8 100644
--- a/net/ceph/osd_client.c
+++ b/net/ceph/osd_client.c
@@ -5060,6 +5060,7 @@ int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
 	struct ceph_osd_request *req;
 	struct page **pages;
 	int ret;
+	const u32 buf_len = PAGE_SIZE;
 
 	req = ceph_osdc_alloc_request(osdc, NULL, 1, false, GFP_NOIO);
 	if (!req)
@@ -5078,7 +5079,7 @@ int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
 	osd_req_op_init(req, 0, CEPH_OSD_OP_LIST_WATCHERS, 0);
 	ceph_osd_data_pages_init(osd_req_op_data(req, 0, list_watchers,
 						 response_data),
-				 pages, PAGE_SIZE, 0, false, true);
+				 pages, buf_len, 0, false, true);
 
 	ret = ceph_osdc_alloc_messages(req, GFP_NOIO);
 	if (ret)
@@ -5088,7 +5089,8 @@ int ceph_osdc_list_watchers(struct ceph_osd_client *osdc,
 	ret = ceph_osdc_wait_request(osdc, req);
 	if (ret >= 0) {
 		void *p = page_address(pages[0]);
-		void *const end = p + req->r_ops[0].outdata_len;
+		void *const end = p +
+			min_t(u32, req->r_ops[0].outdata_len, buf_len);
 
 		ret = decode_watchers(&p, end, watchers, num_watchers);
 	}
-- 
2.53.0