summaryrefslogtreecommitdiff
path: root/arch/powerpc/kernel/exceptions-64s.S
diff options
context:
space:
mode:
authorMichael Ellerman <mpe@ellerman.id.au>2020-07-23 17:43:44 +1000
committerMichael Ellerman <mpe@ellerman.id.au>2020-07-23 17:43:44 +1000
commit335aca5f65f1a39670944930131da5f2276f888f (patch)
treec8f41223681fc064f558083a2d46f712e43d8b66 /arch/powerpc/kernel/exceptions-64s.S
parent8ac9b9d61f0eceba6ce571e7527798465ae9a7c5 (diff)
parent7fa95f9adaee7e5cbb195d3359741120829e488b (diff)
downloadlinux-next-335aca5f65f1a39670944930131da5f2276f888f.tar.gz
linux-next-335aca5f65f1a39670944930131da5f2276f888f.zip
Merge branch 'scv' support into next
From Nick's cover letter: Linux powerpc new system call instruction and ABI System Call Vectored (scv) ABI ============================== The scv instruction is introduced with POWER9 / ISA3, it comes with an rfscv counter-part. The benefit of these instructions is performance (trading slower SRR0/1 with faster LR/CTR registers, and entering the kernel with MSR[EE] and MSR[RI] left enabled, which can reduce MSR updates. The scv instruction has 128 levels (not enough to cover the Linux system call space). Assignment and advertisement ---------------------------- The proposal is to assign scv levels conservatively, and advertise them with HWCAP feature bits as we add support for more. Linux has not enabled FSCR[SCV] yet, so executing the scv instruction will cause the kernel to log a "SCV facility unavilable" message, and deliver a SIGILL with ILL_ILLOPC to the process. Linux has defined a HWCAP2 bit PPC_FEATURE2_SCV for SCV support, but does not set it. This change allocates the zero level ('scv 0'), advertised with PPC_FEATURE2_SCV, which will be used to provide normal Linux system calls (equivalent to 'sc'). Attempting to execute scv with other levels will cause a SIGILL to be delivered the same as before, but will not log a "SCV facility unavailable" message (because the processor facility is enabled). Calling convention ------------------ The proposal is for scv 0 to provide the standard Linux system call ABI with the following differences from sc convention[1]: - LR is to be volatile across scv calls. This is necessary because the scv instruction clobbers LR. From previous discussion, this should be possible to deal with in GCC clobbers and CFI. - cr1 and cr5-cr7 are volatile. This matches the C ABI and would allow the kernel system call exit to avoid restoring the volatile cr registers (although we probably still would anyway to avoid information leaks). - Error handling: The consensus among kernel, glibc, and musl is to move to using negative return values in r3 rather than CR0[SO]=1 to indicate error, which matches most other architectures, and is closer to a function call. Notes ----- - r0,r4-r8 are documented as volatile in the ABI, but the kernel patch as submitted currently preserves them. This is to leave room for deciding which way to go with these. Some small benefit was found by preserving them[1] but I'm not convinced it's worth deviating from the C function call ABI just for this. Release code should follow the ABI. Previous discussions: https://lists.ozlabs.org/pipermail/linuxppc-dev/2020-April/208691.html https://lists.ozlabs.org/pipermail/linuxppc-dev/2020-April/209268.html [1] https://github.com/torvalds/linux/blob/master/Documentation/powerpc/syscall64-abi.rst [2] https://lists.ozlabs.org/pipermail/linuxppc-dev/2020-April/209263.html
Diffstat (limited to 'arch/powerpc/kernel/exceptions-64s.S')
-rw-r--r--arch/powerpc/kernel/exceptions-64s.S150
1 files changed, 143 insertions, 7 deletions
diff --git a/arch/powerpc/kernel/exceptions-64s.S b/arch/powerpc/kernel/exceptions-64s.S
index 0fc8bad878b2..eda5ceef2381 100644
--- a/arch/powerpc/kernel/exceptions-64s.S
+++ b/arch/powerpc/kernel/exceptions-64s.S
@@ -508,8 +508,24 @@ DEFINE_FIXED_SYMBOL(\name\()_common_real)
.macro __GEN_COMMON_BODY name
.if IMASK
+ .if ! ISTACK
+ .error "No support for masked interrupt to use custom stack"
+ .endif
+
+ /* If coming from user, skip soft-mask tests. */
+ andi. r10,r12,MSR_PR
+ bne 2f
+
+ /* Kernel code running below __end_interrupts is implicitly
+ * soft-masked */
+ LOAD_HANDLER(r10, __end_interrupts)
+ cmpld r11,r10
+ li r10,IMASK
+ blt- 1f
+
+ /* Test the soft mask state against our interrupt's bit */
lbz r10,PACAIRQSOFTMASK(r13)
- andi. r10,r10,IMASK
+1: andi. r10,r10,IMASK
/* Associate vector numbers with bits in paca->irq_happened */
.if IVEC == 0x500 || IVEC == 0xea0
li r10,PACA_IRQ_EE
@@ -540,7 +556,7 @@ DEFINE_FIXED_SYMBOL(\name\()_common_real)
.if ISTACK
andi. r10,r12,MSR_PR /* See if coming from user */
- mr r10,r1 /* Save r1 */
+2: mr r10,r1 /* Save r1 */
subi r1,r1,INT_FRAME_SIZE /* alloc frame on kernel stack */
beq- 100f
ld r1,PACAKSAVE(r13) /* kernel stack to use */
@@ -740,6 +756,9 @@ END_FTR_SECTION_IFSET(CPU_FTR_CAN_NAP)
* guarantee they will be delivered virtually. Some conditions (see the ISA)
* cause exceptions to be delivered in real mode.
*
+ * The scv instructions are a special case. They get a 0x3000 offset applied.
+ * scv exceptions have unique reentrancy properties, see below.
+ *
* It's impossible to receive interrupts below 0x300 via AIL.
*
* KVM: None of the virtual exceptions are from the guest. Anything that
@@ -749,8 +768,8 @@ END_FTR_SECTION_IFSET(CPU_FTR_CAN_NAP)
* We layout physical memory as follows:
* 0x0000 - 0x00ff : Secondary processor spin code
* 0x0100 - 0x18ff : Real mode pSeries interrupt vectors
- * 0x1900 - 0x3fff : Real mode trampolines
- * 0x4000 - 0x58ff : Relon (IR=1,DR=1) mode pSeries interrupt vectors
+ * 0x1900 - 0x2fff : Real mode trampolines
+ * 0x3000 - 0x58ff : Relon (IR=1,DR=1) mode pSeries interrupt vectors
* 0x5900 - 0x6fff : Relon mode trampolines
* 0x7000 - 0x7fff : FWNMI data area
* 0x8000 - .... : Common interrupt handlers, remaining early
@@ -761,8 +780,8 @@ END_FTR_SECTION_IFSET(CPU_FTR_CAN_NAP)
* vectors there.
*/
OPEN_FIXED_SECTION(real_vectors, 0x0100, 0x1900)
-OPEN_FIXED_SECTION(real_trampolines, 0x1900, 0x4000)
-OPEN_FIXED_SECTION(virt_vectors, 0x4000, 0x5900)
+OPEN_FIXED_SECTION(real_trampolines, 0x1900, 0x3000)
+OPEN_FIXED_SECTION(virt_vectors, 0x3000, 0x5900)
OPEN_FIXED_SECTION(virt_trampolines, 0x5900, 0x7000)
#ifdef CONFIG_PPC_POWERNV
@@ -798,6 +817,77 @@ USE_FIXED_SECTION(real_vectors)
.globl __start_interrupts
__start_interrupts:
+/**
+ * Interrupt 0x3000 - System Call Vectored Interrupt (syscall).
+ * This is a synchronous interrupt invoked with the "scv" instruction. The
+ * system call does not alter the HV bit, so it is directed to the OS.
+ *
+ * Handling:
+ * scv instructions enter the kernel without changing EE, RI, ME, or HV.
+ * In particular, this means we can take a maskable interrupt at any point
+ * in the scv handler, which is unlike any other interrupt. This is solved
+ * by treating the instruction addresses below __end_interrupts as being
+ * soft-masked.
+ *
+ * AIL-0 mode scv exceptions go to 0x17000-0x17fff, but we set AIL-3 and
+ * ensure scv is never executed with relocation off, which means AIL-0
+ * should never happen.
+ *
+ * Before leaving the below __end_interrupts text, at least of the following
+ * must be true:
+ * - MSR[PR]=1 (i.e., return to userspace)
+ * - MSR_EE|MSR_RI is set (no reentrant exceptions)
+ * - Standard kernel environment is set up (stack, paca, etc)
+ *
+ * Call convention:
+ *
+ * syscall register convention is in Documentation/powerpc/syscall64-abi.rst
+ */
+EXC_VIRT_BEGIN(system_call_vectored, 0x3000, 0x1000)
+ /* SCV 0 */
+ mr r9,r13
+ GET_PACA(r13)
+ mflr r11
+ mfctr r12
+ li r10,IRQS_ALL_DISABLED
+ stb r10,PACAIRQSOFTMASK(r13)
+#ifdef CONFIG_RELOCATABLE
+ b system_call_vectored_tramp
+#else
+ b system_call_vectored_common
+#endif
+ nop
+
+ /* SCV 1 - 127 */
+ .rept 127
+ mr r9,r13
+ GET_PACA(r13)
+ mflr r11
+ mfctr r12
+ li r10,IRQS_ALL_DISABLED
+ stb r10,PACAIRQSOFTMASK(r13)
+ li r0,-1 /* cause failure */
+#ifdef CONFIG_RELOCATABLE
+ b system_call_vectored_sigill_tramp
+#else
+ b system_call_vectored_sigill
+#endif
+ .endr
+EXC_VIRT_END(system_call_vectored, 0x3000, 0x1000)
+
+#ifdef CONFIG_RELOCATABLE
+TRAMP_VIRT_BEGIN(system_call_vectored_tramp)
+ __LOAD_HANDLER(r10, system_call_vectored_common)
+ mtctr r10
+ bctr
+
+TRAMP_VIRT_BEGIN(system_call_vectored_sigill_tramp)
+ __LOAD_HANDLER(r10, system_call_vectored_sigill)
+ mtctr r10
+ bctr
+#endif
+
+
/* No virt vectors corresponding with 0x0..0x100 */
EXC_VIRT_NONE(0x4000, 0x100)
@@ -2838,7 +2928,8 @@ masked_interrupt:
ld r10,PACA_EXGEN+EX_R10(r13)
ld r11,PACA_EXGEN+EX_R11(r13)
ld r12,PACA_EXGEN+EX_R12(r13)
- /* returns to kernel where r13 must be set up, so don't restore it */
+ ld r13,PACA_EXGEN+EX_R13(r13)
+ /* May return to masked low address where r13 is not set up */
.if \hsrr
HRFI_TO_KERNEL
.else
@@ -2946,6 +3037,47 @@ TRAMP_REAL_BEGIN(hrfi_flush_fallback)
GET_SCRATCH0(r13);
hrfid
+TRAMP_REAL_BEGIN(rfscv_flush_fallback)
+ /* system call volatile */
+ mr r7,r13
+ GET_PACA(r13);
+ mr r8,r1
+ ld r1,PACAKSAVE(r13)
+ mfctr r9
+ ld r10,PACA_RFI_FLUSH_FALLBACK_AREA(r13)
+ ld r11,PACA_L1D_FLUSH_SIZE(r13)
+ srdi r11,r11,(7 + 3) /* 128 byte lines, unrolled 8x */
+ mtctr r11
+ DCBT_BOOK3S_STOP_ALL_STREAM_IDS(r11) /* Stop prefetch streams */
+
+ /* order ld/st prior to dcbt stop all streams with flushing */
+ sync
+
+ /*
+ * The load adresses are at staggered offsets within cachelines,
+ * which suits some pipelines better (on others it should not
+ * hurt).
+ */
+1:
+ ld r11,(0x80 + 8)*0(r10)
+ ld r11,(0x80 + 8)*1(r10)
+ ld r11,(0x80 + 8)*2(r10)
+ ld r11,(0x80 + 8)*3(r10)
+ ld r11,(0x80 + 8)*4(r10)
+ ld r11,(0x80 + 8)*5(r10)
+ ld r11,(0x80 + 8)*6(r10)
+ ld r11,(0x80 + 8)*7(r10)
+ addi r10,r10,0x80*8
+ bdnz 1b
+
+ mtctr r9
+ li r9,0
+ li r10,0
+ li r11,0
+ mr r1,r8
+ mr r13,r7
+ RFSCV
+
USE_TEXT_SECTION()
MASKED_INTERRUPT
MASKED_INTERRUPT hsrr=1
@@ -2997,6 +3129,10 @@ EXC_COMMON_BEGIN(ppc64_runlatch_on_trampoline)
USE_FIXED_SECTION(virt_trampolines)
/*
+ * All code below __end_interrupts is treated as soft-masked. If
+ * any code runs here with MSR[EE]=1, it must then cope with pending
+ * soft interrupt being raised (i.e., by ensuring it is replayed).
+ *
* The __end_interrupts marker must be past the out-of-line (OOL)
* handlers, so that they are copied to real address 0x100 when running
* a relocatable kernel. This ensures they can be reached from the short