On Thu, Mar 21, 2019 at 10:45:31AM -0700, Kees Cook wrote:

On Wed, Mar 20, 2019 at 11:57 AM Eric Biggers [off-list ref] wrote:

quoted

On Tue, Mar 19, 2019 at 10:09:13AM -0700, Eric Biggers wrote:

quoted

On Tue, Mar 19, 2019 at 12:54:23PM +0100, Geert Uytterhoeven wrote:

quoted

When running the sha1-asm crypto selftest on arm with
CONFIG_HARDENED_USERCOPY_PAGESPAN=y:

    usercopy: Kernel memory overwrite attempt detected to spans
multiple pages (offset 0, size 42)!
    ------------[ cut here ]------------
    kernel BUG at mm/usercopy.c:102!
    Internal error: Oops - BUG: 0 [#1] SMP ARM
    Modules linked in:
    CPU: 0 PID: 35 Comm: cryptomgr_test Not tainted
5.1.0-rc1-koelsch-01109-gbeb7d6376ecfbf07-dirty #397
    Hardware name: Generic R-Car Gen2 (Flattened Device Tree)
    PC is at usercopy_abort+0x68/0x90
    LR is at usercopy_abort+0x68/0x90
    pc : [<c030fd60>]    lr : [<c030fd60>]    psr: 60000013
    sp : ea54bc60  ip : 00000010  fp : cccccccd
    r10: 00000000  r9 : c0e0ce04  r8 : ea54d009
    r7 : ea54d00a  r6 : 00000000  r5 : 0000002a  r4 : c09d1120
    r3 : dd6cd422  r2 : dd6cd422  r1 : 2abb4000  r0 : 0000005f
    Flags: nZCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment user
    Control: 30c5387d  Table: 40003000  DAC: fffffffd
    Process cryptomgr_test (pid: 35, stack limit = 0x(ptrval))
    Stack: (0xea54bc60 to 0xea54c000)
    bc60: c09d1120 c09d1120 c09d1120 00000000 0000002a 0000002a
00000000 c0310060
    bc80: 0000002a 00000000 000001c0 00000000 00000000 c0eb11e8
ea54cfe0 ea538c00
    bca0: 00000000 ea54cfe0 ebef73e0 0000002a ea538c20 ea54bd84
0000003a c0427a30
    bcc0: ea54bdbc 00000000 00000000 c081cf70 eb074280 c081cf70
0000002a c081cf80
    bce0: 0000000e c07da138 ea54bd0c 00000000 c084061c c04248e8
c0e0a408 eb074240
    bd00: eb074200 c04253c8 eb074280 ea550000 00000012 dd6cd422
ebef7480 eb074200
    bd20: ea54bd84 c081cf64 ea537200 00000002 00000000 00000014
c084061c c0428c38
    bd40: ea54bd84 ea54bdbc c081cd34 00000000 c0e4e4b4 ea538c40
00000002 eabe4e80
    bd60: ea538c00 00000400 ea4f7a00 ea4f7a60 eb074240 00000060
00000006 c09d544c
    bd80: 00000038 00000003 00000000 00000038 ea54bd7c 00000001
eb074200 00000000
    bda0: 00000000 dead4ead ffffffff ffffffff ea54bdb0 ea54bdb0
00000000 c081cf70
    bdc0: c081ce68 c081ce78 ea4f7480 eb000780 00000dc0 eb000780
c0e4ee80 443e9884
    bde0: 6ed23b1c a14aaeba e52951f9 f17046e5 fefefefe fefefefe
fefefefe fefefefe
    be00: eb000780 c04292c4 c0e0a638 60000013 60000013 c0305298
ea4f7a00 c03062bc
    be20: eb000780 00000cc0 ea4f7a00 dd6cd422 00000cc0 ea538c00
00000002 eabe4e40
    be40: ea537200 00000007 00000000 ea4f7a00 eb074200 c0429314
eb074200 ea538c00
    be60: ea4f7a00 0000000a eabe4e80 c084061c c08405fc 00000006
c04dace8 00000006
    be80: 00000000 c084065c ea537200 0000000e 00000400 eb04de08
ea4f71a8 c0429420
    bea0: 00000400 ea537200 0000000e ea537200 0000000e c0429374
00000400 ffffffff
    bec0: 000000a2 c042a414 00000103 c0e0a408 00000000 c0e0a438
c0e5a2a0 c0e5a2a0
    bee0: 00000001 00000001 00000017 ffffe000 00000000 60000013
c0e5a2a0 c0269470
    bf00: c09c9ed0 ea54bf5c 00000103 00000000 00000000 c0e0a408
ea537280 0000000e
    bf20: 00000400 c0426500 00000000 eb04de08 ea4f71a8 c02694f4
c09c9ed0 ea54bf5c
    bf40: ea54bf28 c02699d0 ea54bf5c dd6cd422 ea537200 dd6cd422
c09c9ed0 ea537200
    bf60: ea4af1c0 ea54a000 ea537200 c0426500 00000000 eb04de08
ea4f71a8 c0426524
    bf80: ea4f7180 c023dcec ea54a000 ea4af1c0 c023dbb4 00000000
00000000 00000000
    bfa0: 00000000 00000000 00000000 c02010d8 00000000 00000000
00000000 00000000
    bfc0: 00000000 00000000 00000000 00000000 00000000 00000000
00000000 00000000
    bfe0: 00000000 00000000 00000000 00000000 00000013 00000000
00000000 00000000
    [<c030fd60>] (usercopy_abort) from [<c0310060>]
(__check_object_size+0x2d8/0x448)
    [<c0310060>] (__check_object_size) from [<c0427a30>]
(build_test_sglist+0x268/0x2d8)
    [<c0427a30>] (build_test_sglist) from [<c0428c38>]
(test_hash_vec_cfg+0x110/0x694)
    [<c0428c38>] (test_hash_vec_cfg) from [<c0429314>]
(__alg_test_hash+0x158/0x1b8)
    [<c0429314>] (__alg_test_hash) from [<c0429420>] (alg_test_hash+0xac/0xf4)
    [<c0429420>] (alg_test_hash) from [<c042a414>] (alg_test.part.4+0x264/0x2f8)
    [<c042a414>] (alg_test.part.4) from [<c0426524>] (cryptomgr_test+0x24/0x44)
    [<c0426524>] (cryptomgr_test) from [<c023dcec>] (kthread+0x138/0x150)
    [<c023dcec>] (kthread) from [<c02010d8>] (ret_from_fork+0x14/0x3c)
    Exception stack(0xea54bfb0 to 0xea54bff8)
    bfa0:                                     00000000 00000000
00000000 00000000
    bfc0: 00000000 00000000 00000000 00000000 00000000 00000000
00000000 00000000
    bfe0: 00000000 00000000 00000000 00000000 00000013 00000000
    Code: e58de000 e98d0012 e1a0100c ebfd6712 (e7f001f2)
    ---[ end trace 190b3cf48e720f78 ]---
    BUG: sleeping function called from invalid context at
include/linux/percpu-rwsem.h:34
    in_atomic(): 0, irqs_disabled(): 128, pid: 35, name: cryptomgr_test
    CPU: 0 PID: 35 Comm: cryptomgr_test Tainted: G      D
5.1.0-rc1-koelsch-01109-gbeb7d6376ecfbf07-dirty #397
    Hardware name: Generic R-Car Gen2 (Flattened Device Tree)
    [<c020ec74>] (unwind_backtrace) from [<c020ae58>] (show_stack+0x10/0x14)
    [<c020ae58>] (show_stack) from [<c07c3624>] (dump_stack+0x7c/0x9c)
    [<c07c3624>] (dump_stack) from [<c0242e14>] (___might_sleep+0xf4/0x158)
    [<c0242e14>] (___might_sleep) from [<c0230210>] (exit_signals+0x2c/0x258)
    [<c0230210>] (exit_signals) from [<c0223d6c>] (do_exit+0x114/0xa20)
    [<c0223d6c>] (do_exit) from [<c020b160>] (die+0x304/0x344)
    [<c020b160>] (die) from [<c020b388>] (do_undefinstr+0x80/0x190)
    [<c020b388>] (do_undefinstr) from [<c0201b24>] (__und_svc_finish+0x0/0x3c)
    Exception stack(0xea54bc10 to 0xea54bc58)
    bc00:                                     0000005f 2abb4000
dd6cd422 dd6cd422
    bc20: c09d1120 0000002a 00000000 ea54d00a ea54d009 c0e0ce04
00000000 cccccccd
    bc40: 00000010 ea54bc60 c030fd60 c030fd60 60000013 ffffffff
    [<c0201b24>] (__und_svc_finish) from [<c030fd60>] (usercopy_abort+0x68/0x90)
    [<c030fd60>] (usercopy_abort) from [<c0310060>]
(__check_object_size+0x2d8/0x448)
    [<c0310060>] (__check_object_size) from [<c0427a30>]
(build_test_sglist+0x268/0x2d8)
    [<c0427a30>] (build_test_sglist) from [<c0428c38>]
(test_hash_vec_cfg+0x110/0x694)
    [<c0428c38>] (test_hash_vec_cfg) from [<c0429314>]
(__alg_test_hash+0x158/0x1b8)
    [<c0429314>] (__alg_test_hash) from [<c0429420>] (alg_test_hash+0xac/0xf4)
    [<c0429420>] (alg_test_hash) from [<c042a414>] (alg_test.part.4+0x264/0x2f8)
    [<c042a414>] (alg_test.part.4) from [<c0426524>] (cryptomgr_test+0x24/0x44)
    [<c0426524>] (cryptomgr_test) from [<c023dcec>] (kthread+0x138/0x150)
    [<c023dcec>] (kthread) from [<c02010d8>] (ret_from_fork+0x14/0x3c)
    Exception stack(0xea54bfb0 to 0xea54bff8)
    bfa0:                                     00000000 00000000
00000000 00000000
    bfc0: 00000000 00000000 00000000 00000000 00000000 00000000
00000000 00000000
    bfe0: 00000000 00000000 00000000 00000000 00000013 00000000

Well, this must happen with the new (in 5.1) crypto self-tests implementation
for any crypto algorithm when CONFIG_HARDENED_USERCOPY_PAGESPAN=y.  I don't
understand why hardened usercopy considers it a bug though, as there's no buffer
overflow.  The crypto tests use copy_from_iter() to copy data into a 2-page
buffer that was allocated with __get_free_pages():

      __get_free_pages(GFP_KERNEL, 1)

... where 1 means an order-1 allocation.

If it copies to offset=4064 len=42, for example, then hardened usercopy
considers it a bug even though the buffer is 8192 bytes long.  Why?

It isn't actually copying anything to/from userspace, BTW; it's using iov_iter
with ITER_KVEC.

- Eric

Kees, any thoughts on why hardened usercopy rejects copies spanning a page
boundary when they seem to be fine?

This is due to missing the compound page marking, if I remember
correctly. However, I tend to leave the pagespan test disabled: it
really isn't ready for production use -- there are a lot of missing
annotations still.

So do I need to add __GFP_COMP?  Is there any actual reason to do so?
Why does hardened usercopy check for it?

- Eric