riscv: VMAP_STACK overflow detection thread-safe

commit 31da94c25aea ("riscv: add VMAP_STACK overflow detection") added
support for CONFIG_VMAP_STACK. If overflow is detected, CPU switches to
`shadow_stack` temporarily before switching finally to per-cpu
`overflow_stack`.

If two CPUs/harts are racing and end up in over flowing kernel stack, one
or both will end up corrupting each other state because `shadow_stack` is
not per-cpu. This patch optimizes per-cpu overflow stack switch by
directly picking per-cpu `overflow_stack` and gets rid of `shadow_stack`.

Following are the changes in this patch

 - Defines an asm macro to obtain per-cpu symbols in destination
   register.
 - In entry.S, when overflow is detected, per-cpu overflow stack is
   located using per-cpu asm macro. Computing per-cpu symbol requires
   a temporary register. x31 is saved away into CSR_SCRATCH
   (CSR_SCRATCH is anyways zero since we're in kernel).

Please see Links for additional relevant disccussion and alternative
solution.

Tested by `echo EXHAUST_STACK > /sys/kernel/debug/provoke-crash/DIRECT`
Kernel crash log below

 Insufficient stack space to handle exception!/debug/provoke-crash/DIRECT
 Task stack:     [0xff20000010a98000..0xff20000010a9c000]
 Overflow stack: [0xff600001f7d98370..0xff600001f7d99370]
 CPU: 1 PID: 205 Comm: bash Not tainted 6.1.0-rc2-00001-g328a1f96f7b9 #34
 Hardware name: riscv-virtio,qemu (DT)
 epc : __memset+0x60/0xfc
  ra : recursive_loop+0x48/0xc6 [lkdtm]
 epc : ffffffff808de0e4 ra : ffffffff0163a752 sp : ff20000010a97e80
  gp : ffffffff815c0330 tp : ff600000820ea280 t0 : ff20000010a97e88
  t1 : 000000000000002e t2 : 3233206874706564 s0 : ff20000010a982b0
  s1 : 0000000000000012 a0 : ff20000010a97e88 a1 : 0000000000000000
  a2 : 0000000000000400 a3 : ff20000010a98288 a4 : 0000000000000000
  a5 : 0000000000000000 a6 : fffffffffffe43f0 a7 : 00007fffffffffff
  s2 : ff20000010a97e88 s3 : ffffffff01644680 s4 : ff20000010a9be90
  s5 : ff600000842ba6c0 s6 : 00aaaaaac29e42b0 s7 : 00fffffff0aa3684
  s8 : 00aaaaaac2978040 s9 : 0000000000000065 s10: 00ffffff8a7cad10
  s11: 00ffffff8a76a4e0 t3 : ffffffff815dbaf4 t4 : ffffffff815dbaf4
  t5 : ffffffff815dbab8 t6 : ff20000010a9bb48
 status: 0000000200000120 badaddr: ff20000010a97e88 cause: 000000000000000f
 Kernel panic - not syncing: Kernel stack overflow
 CPU: 1 PID: 205 Comm: bash Not tainted 6.1.0-rc2-00001-g328a1f96f7b9 #34
 Hardware name: riscv-virtio,qemu (DT)
 Call Trace:
 [<ffffffff80006754>] dump_backtrace+0x30/0x38
 [<ffffffff808de798>] show_stack+0x40/0x4c
 [<ffffffff808ea2a8>] dump_stack_lvl+0x44/0x5c
 [<ffffffff808ea2d8>] dump_stack+0x18/0x20
 [<ffffffff808dec06>] panic+0x126/0x2fe
 [<ffffffff800065ea>] walk_stackframe+0x0/0xf0
 [<ffffffff0163a752>] recursive_loop+0x48/0xc6 [lkdtm]
 SMP: stopping secondary CPUs
 ---[ end Kernel panic - not syncing: Kernel stack overflow ]---

Cc: Guo Ren <guoren@kernel.org>
Cc: Jisheng Zhang <jszhang@kernel.org>
Link: https://lore.kernel.org/linux-riscv/Y347B0x4VUNOd6V7@xhacker/T/#t
Link: https://lore.kernel.org/lkml/20221124094845.1907443-1-debug@rivosinc.com/
Signed-off-by: Deepak Gupta <debug@rivosinc.com>
Co-developed-by: Sami Tolvanen <samitolvanen@google.com>
Signed-off-by: Sami Tolvanen <samitolvanen@google.com>
Acked-by: Guo Ren <guoren@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Link: https://lore.kernel.org/r/20230927224757.1154247-9-samitolvanen@google.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>

1 files changed, 10 insertions, 60 deletions

diff --git a/arch/riscv/kernel/entry.S b/arch/riscv/kernel/entry.S
index 143a2bb3e697..3d11aa3af105 100644
--- a/arch/riscv/kernel/entry.S
+++ b/arch/riscv/kernel/entry.S
@@ -10,9 +10,11 @@
 #include <asm/asm.h>
 #include <asm/csr.h>
 #include <asm/unistd.h>
+#include <asm/page.h>
 #include <asm/thread_info.h>
 #include <asm/asm-offsets.h>
 #include <asm/errata_list.h>
+#include <linux/sizes.h>
 
 SYM_CODE_START(handle_exception)
 	/*
@@ -170,67 +172,15 @@ SYM_CODE_END(ret_from_exception)
 
 #ifdef CONFIG_VMAP_STACK
 SYM_CODE_START_LOCAL(handle_kernel_stack_overflow)
-	/*
-	 * Takes the psuedo-spinlock for the shadow stack, in case multiple
-	 * harts are concurrently overflowing their kernel stacks.  We could
-	 * store any value here, but since we're overflowing the kernel stack
-	 * already we only have SP to use as a scratch register.  So we just
-	 * swap in the address of the spinlock, as that's definately non-zero.
-	 *
-	 * Pairs with a store_release in handle_bad_stack().
-	 */
-1:	la sp, spin_shadow_stack
-	REG_AMOSWAP_AQ sp, sp, (sp)
-	bnez sp, 1b
-
-	la sp, shadow_stack
-	addi sp, sp, SHADOW_OVERFLOW_STACK_SIZE
-
-	//save caller register to shadow stack
-	addi sp, sp, -(PT_SIZE_ON_STACK)
-	REG_S x1,  PT_RA(sp)
-	REG_S x5,  PT_T0(sp)
-	REG_S x6,  PT_T1(sp)
-	REG_S x7,  PT_T2(sp)
-	REG_S x10, PT_A0(sp)
-	REG_S x11, PT_A1(sp)
-	REG_S x12, PT_A2(sp)
-	REG_S x13, PT_A3(sp)
-	REG_S x14, PT_A4(sp)
-	REG_S x15, PT_A5(sp)
-	REG_S x16, PT_A6(sp)
-	REG_S x17, PT_A7(sp)
-	REG_S x28, PT_T3(sp)
-	REG_S x29, PT_T4(sp)
-	REG_S x30, PT_T5(sp)
-	REG_S x31, PT_T6(sp)
-
-	la ra, restore_caller_reg
-	tail get_overflow_stack
-
-restore_caller_reg:
-	//save per-cpu overflow stack
-	REG_S a0, -8(sp)
-	//restore caller register from shadow_stack
-	REG_L x1,  PT_RA(sp)
-	REG_L x5,  PT_T0(sp)
-	REG_L x6,  PT_T1(sp)
-	REG_L x7,  PT_T2(sp)
-	REG_L x10, PT_A0(sp)
-	REG_L x11, PT_A1(sp)
-	REG_L x12, PT_A2(sp)
-	REG_L x13, PT_A3(sp)
-	REG_L x14, PT_A4(sp)
-	REG_L x15, PT_A5(sp)
-	REG_L x16, PT_A6(sp)
-	REG_L x17, PT_A7(sp)
-	REG_L x28, PT_T3(sp)
-	REG_L x29, PT_T4(sp)
-	REG_L x30, PT_T5(sp)
-	REG_L x31, PT_T6(sp)
+	/* we reach here from kernel context, sscratch must be 0 */
+	csrrw x31, CSR_SCRATCH, x31
+	asm_per_cpu sp, overflow_stack, x31
+	li x31, OVERFLOW_STACK_SIZE
+	add sp, sp, x31
+	/* zero out x31 again and restore x31 */
+	xor x31, x31, x31
+	csrrw x31, CSR_SCRATCH, x31
 
-	//load per-cpu overflow stack
-	REG_L sp, -8(sp)
 	addi sp, sp, -(PT_SIZE_ON_STACK)
 
 	//save context to overflow stack