Re: [PATCH] riscv: Fixup bootup failure with HARDENED_USERCOPY
From: Palmer Dabbelt
Date: Wed Oct 07 2020 - 00:15:30 EST
On Tue, 06 Oct 2020 09:49:33 PDT (-0700), guoren@xxxxxxxxxx wrote:
From: Guo Ren <guoren@xxxxxxxxxxxxxxxxx>
As Aurelien has reported:
[ 3.484586] AppArmor: AppArmor sha1 policy hashing enabled
[ 4.749835] Freeing unused kernel memory: 492K
[ 4.752017] Run /init as init process
[ 4.753571] usercopy: Kernel memory overwrite attempt detected to kernel text (offset 507879, size 11)!
[ 4.754838] ------------[ cut here ]------------
[ 4.755651] kernel BUG at mm/usercopy.c:99!
[ 4.756445] Kernel BUG [#1]
[ 4.756815] Modules linked in:
[ 4.757542] CPU: 1 PID: 1 Comm: swapper/0 Not tainted 5.8.0-1-riscv64 #1 Debian 5.8.7-1
[ 4.758372] epc: ffffffe0003b5120 ra : ffffffe0003b5120 sp : ffffffe07f783ca0
[ 4.758960] gp : ffffffe000cc7230 tp : ffffffe07f77cec0 t0 : ffffffe000cdafc0
[ 4.759772] t1 : 0000000000000064 t2 : 0000000000000000 s0 : ffffffe07f783cf0
[ 4.760534] s1 : ffffffe00095d780 a0 : 000000000000005b a1 : 0000000000000020
[ 4.761309] a2 : 0000000000000005 a3 : 0000000000000000 a4 : ffffffe000c1f340
[ 4.761848] a5 : ffffffe000c1f340 a6 : 0000000000000000 a7 : 0000000000000087
[ 4.762684] s2 : ffffffe000941848 s3 : 000000000007bfe7 s4 : 000000000000000b
[ 4.763500] s5 : 0000000000000000 s6 : ffffffe00091cc00 s7 : fffffffffffff000
[ 4.764376] s8 : 0000003ffffff000 s9 : ffffffe0769f3200 s10: 000000000000000b
[ 4.765208] s11: ffffffe07d548c40 t3 : 0000000000000000 t4 : 000000000001dcd0
[ 4.766059] t5 : ffffffe000cc8510 t6 : ffffffe000cd64aa
[ 4.766712] status: 0000000000000120 badaddr: 0000000000000000 cause: 0000000000000003
[ 4.768308] ---[ end trace 1f8e733e834d4c3e ]---
[ 4.769129] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b
[ 4.770070] SMP: stopping secondary CPUs
[ 4.771110] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ]---
Above failure is relate to commit: a0fa4027dc911 (riscv: Fixup
That commit isn't in Linus' tree (at least, as far as I see it). I have
6184358da000 ("riscv: Fixup static_obj() fail"), so I'm going to fix that -- in
fact, I'm going to essentially just replace most of this rationale with what I
wrote up in my revert as this is all a bit too long for a commit message.
I was kind of worried the initdata move was a bit too risky, but after reading
the users of __init_{begin,end} I think it's safe. Here's what I ended up with
on fixes.
Thanks!
commit 84814460eef9af0fb56a4698341c9cb7996a6312 (HEAD -> fixes, riscv/fixes)
gpg: Signature made Tue 06 Oct 2020 09:11:35 PM PDT
gpg: using RSA key 2B3C3747446843B24A943A7A2E1319F35FBB1889
gpg: issuer "palmer@xxxxxxxxxxx"
gpg: Good signature from "Palmer Dabbelt <palmer@xxxxxxxxxxx>" [ultimate]
gpg: aka "Palmer Dabbelt <palmerdabbelt@xxxxxxxxxx>" [ultimate]
Author: Guo Ren <guoren@xxxxxxxxxxxxxxxxx>
Date: Tue Oct 6 16:49:33 2020 +0000
riscv: Fixup bootup failure with HARDENED_USERCOPY
6184358da000 ("riscv: Fixup static_obj() fail") attempted to elide a lockdep
failure by rearranging our kernel image to place all initdata within [_stext,
_end], thus triggering lockdep to treat these as static objects. These objects
are released and eventually reallocated, causing check_kernel_text_object() to
trigger a BUG().
This backs out the change to make [_stext, _end] all-encompassing, instead just
moving initdata. This results in initdata being outside of [__init_begin,
__init_end], which means initdata can't be freed.
Link: https://lore.kernel.org/linux-riscv/1593266228-61125-1-git-send-email-guoren@xxxxxxxxxx/T/#t
Signed-off-by: Guo Ren <guoren@xxxxxxxxxxxxxxxxx>
Reported-by: Aurelien Jarno <aurelien@xxxxxxxxxxx>
Tested-by: Aurelien Jarno <aurelien@xxxxxxxxxxx>
[Palmer: Clean up commit text]
Signed-off-by: Palmer Dabbelt <palmerdabbelt@xxxxxxxxxx>
static_obj() fail). When we expand static_obj include INIT_DATA,
we also include INIT_TEXT into usercopy check kernel text:
/* Is this address range in the kernel text area? */
static inline void check_kernel_text_object(const unsigned long ptr,
unsigned long n, bool to_user)
{
unsigned long textlow = (unsigned long)_stext;
unsigned long texthigh = (unsigned long)_etext;
unsigned long textlow_linear, texthigh_linear;
if (overlaps(ptr, n, textlow, texthigh))
usercopy_abort("kernel text", NULL, to_user, ptr - textlow, n);
When INIT_TEXT/DATA are freed, new allocation will reuse these
memory and overlaps check will be triggered.
The patch met static_obj and check_kernel_text_object requirements.
Link: https://lore.kernel.org/linux-riscv/1593266228-61125-1-git-send-email-guoren@xxxxxxxxxx/T/#t
Signed-off-by: Guo Ren <guoren@xxxxxxxxxxxxxxxxx>
Reported-by: Aurelien Jarno <aurelien@xxxxxxxxxxx>
Tested-by: Aurelien Jarno <aurelien@xxxxxxxxxxx>
Cc: Palmer Dabbelt <palmerdabbelt@xxxxxxxxxx>
Cc: Atish Patra <atishp@xxxxxxxxxxxxxx>
Cc: Andreas Schwab <schwab@xxxxxxxxxxxxxx>
---
arch/riscv/kernel/vmlinux.lds.S | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/arch/riscv/kernel/vmlinux.lds.S b/arch/riscv/kernel/vmlinux.lds.S
index f3586e3..34d00d9 100644
--- a/arch/riscv/kernel/vmlinux.lds.S
+++ b/arch/riscv/kernel/vmlinux.lds.S
@@ -22,13 +22,11 @@ SECTIONS
/* Beginning of code and text segment */
. = LOAD_OFFSET;
_start = .;
- _stext = .;
HEAD_TEXT_SECTION
. = ALIGN(PAGE_SIZE);
__init_begin = .;
INIT_TEXT_SECTION(PAGE_SIZE)
- INIT_DATA_SECTION(16)
. = ALIGN(8);
__soc_early_init_table : {
__soc_early_init_table_start = .;
@@ -55,6 +53,7 @@ SECTIONS
. = ALIGN(SECTION_ALIGN);
.text : {
_text = .;
+ _stext = .;
TEXT_TEXT
SCHED_TEXT
CPUIDLE_TEXT
@@ -67,6 +66,8 @@ SECTIONS
_etext = .;
}
+ INIT_DATA_SECTION(16)
+
/* Start of data section */
_sdata = .;
RO_DATA(SECTION_ALIGN)