Merge tag 'accel-20240426' of https://github.com/philmd/qemu into staging

Accelerators patches

A lot of trivial cleanups and simplifications (moving methods around,
adding/removing #include statements). Most notable changes:

- Rename NEED_CPU_H -> COMPILING_PER_TARGET
- Rename few template headers using the '.h.inc' suffix
- Extract some definitions / declarations into their own header:
  - accel/tcg/user-retaddr.h (helper_retaddr)
  - include/exec/abi_ptr.h (abi_ptr)
  - include/exec/breakpoint.h (CPUBreakpoint, CPUWatchpoint)
  - include/exec/mmu-access-type.h (MMUAccessType)
  - include/user/tswap-target.h (tswapl, bswaptls)

# -----BEGIN PGP SIGNATURE-----
#
# iQIzBAABCAAdFiEE+qvnXhKRciHc/Wuy4+MsLN6twN4FAmYsAuEACgkQ4+MsLN6t
# wN78Rg//V9UoE0U9Lh6Sd2WpcSAYP9D1CBa+iGXhrmel0utER1sQLu022nvcLdHc
# XtCgtX3H0yECF8dPX02rVp8IbSlOv3c8N/a6BxD79cRGqgXBYR/dEUqfXqeLJn3l
# a58YU3i7sLNQ0l7VnwTiBnI0lw170/xJl2B2mcR1SvWuH3dr5vTeIXNureu36ORo
# rc0oqWHbw1Pyyn8ADE2kPyFCOiwPwvcOvAk8dXGfib+mNCwNVV+ZUtAPi711VD8d
# 9VW2gu2sXwnWdpROrSugSw+aPVF4UjltL9qJEl5bxoqWFmlET1Zn2NpKvsocUXmh
# CMQQS2Tr4LpaaVQJGxx0yUe0B65X5+gCkIhsMOubED7GRyTCjrkOPm6exz3ge6WV
# YmIboggFAk3OjAzLs7yZVkWsTK1Y3+3eX0u7AWPUsUu7rCT/Toc6QxDS7eT2hJfq
# UDXI355PGbImgiArQa+OsT7v1Le4/iQa+TfN4fdUDpxEdfaxhnijWh+E91CEp+w/
# Mq7db9Z1aMnhFKIKdkPYyfwB74yXQrmYchJ0QojZjbzqNGwkt9VeC7O9RcYjEaHM
# hMIexwccxexqGH22wn8vPd6ZVKtiLaG4AXO0v6Dn2YJ7/zb/ntcI6lRZqdBAHKNK
# MzkjTRRRR0wAfu8Lk8CaNNEqUP4Po43fbYymo6AZhIR8NqfApL0=
# =ifBx
# -----END PGP SIGNATURE-----
# gpg: Signature made Fri 26 Apr 2024 12:39:13 PM PDT
# gpg:                using RSA key FAABE75E12917221DCFD6BB2E3E32C2CDEADC0DE
# gpg: Good signature from "Philippe Mathieu-Daudé (F4BUG) <f4bug@amsat.org>" [full]

* tag 'accel-20240426' of https://github.com/philmd/qemu: (38 commits)
  plugins: Include missing 'qemu/bitmap.h' header
  hw/core: Avoid including the full 'hw/core/cpu.h' in 'tcg-cpu-ops.h'
  exec: Move CPUTLBEntry helpers to cputlb.c
  exec: Restrict inclusion of 'user/guest-base.h'
  exec: Rename 'exec/user/guest-base.h' as 'user/guest-base.h'
  exec: Restrict 'cpu_ldst.h' to TCG accelerator
  exec: Restrict TCG specific declarations of 'cputlb.h'
  exec: Declare CPUBreakpoint/CPUWatchpoint type in 'breakpoint.h' header
  exec: Declare MMUAccessType type in 'mmu-access-type.h' header
  exec: Declare abi_ptr type in its own 'abi_ptr.h' header
  exec/user: Do not include 'cpu.h' in 'abitypes.h'
  exec: Move [b]tswapl() declarations to 'exec/user/tswap-target.h'
  exec: Declare target_words_bigendian() in 'exec/tswap.h'
  exec/cpu-all: Remove unused tswapls() definitions
  exec/cpu-all: Remove unused 'qemu/thread.h' header
  exec/cpu-all: Reduce 'qemu/rcu.h' header inclusion
  accel/hvf: Use accel-specific per-vcpu @dirty field
  accel/nvmm: Use accel-specific per-vcpu @dirty field
  accel/whpx: Use accel-specific per-vcpu @dirty field
  accel/tcg: Rename helper-head.h -> helper-head.h.inc
  ...

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>

MAINTAINERS

@@ -3692,6 +3692,7 @@ Overall usermode emulation
 M: Riku Voipio <riku.voipio@iki.fi>
 S: Maintained
 F: accel/tcg/user-exec*.c
+F: include/exec/user/
 F: include/user/
 F: common-user/
 

accel/hvf/hvf-accel-ops.c

@@ -204,15 +204,15 @@ static void hvf_set_phys_mem(MemoryRegionSection *section, bool add)
 
 static void do_hvf_cpu_synchronize_state(CPUState *cpu, run_on_cpu_data arg)
 {
-    if (!cpu->vcpu_dirty) {
+    if (!cpu->accel->dirty) {
         hvf_get_registers(cpu);
-        cpu->vcpu_dirty = true;
+        cpu->accel->dirty = true;
     }
 }
 
 static void hvf_cpu_synchronize_state(CPUState *cpu)
 {
-    if (!cpu->vcpu_dirty) {
+    if (!cpu->accel->dirty) {
         run_on_cpu(cpu, do_hvf_cpu_synchronize_state, RUN_ON_CPU_NULL);
     }
 }
@@ -221,7 +221,7 @@ static void do_hvf_cpu_synchronize_set_dirty(CPUState *cpu,
                                              run_on_cpu_data arg)
 {
     /* QEMU state is the reference, push it to HVF now and on next entry */
-    cpu->vcpu_dirty = true;
+    cpu->accel->dirty = true;
 }
 
 static void hvf_cpu_synchronize_post_reset(CPUState *cpu)
@@ -402,7 +402,7 @@ static int hvf_init_vcpu(CPUState *cpu)
 #else
     r = hv_vcpu_create((hv_vcpuid_t *)&cpu->accel->fd, HV_VCPU_DEFAULT);
 #endif
-    cpu->vcpu_dirty = 1;
+    cpu->accel->dirty = true;
     assert_hvf_ok(r);
 
     cpu->accel->guest_debug_enabled = false;

accel/stubs/tcg-stub.c

@@ -18,10 +18,6 @@ void tb_flush(CPUState *cpu)
 {
 }
 
-void tlb_set_dirty(CPUState *cpu, vaddr vaddr)
-{
-}
-
 int probe_access_flags(CPUArchState *env, vaddr addr, int size,
                        MMUAccessType access_type, int mmu_idx,
                        bool nonfault, void **phost, uintptr_t retaddr)

accel/tcg/cpu-exec.c

@@ -41,6 +41,9 @@
 #include "tb-context.h"
 #include "internal-common.h"
 #include "internal-target.h"
+#if defined(CONFIG_USER_ONLY)
+#include "user-retaddr.h"
+#endif
 
 /* -icount align implementation. */
 

accel/tcg/cputlb.c

@@ -27,6 +27,9 @@
 #include "exec/tb-flush.h"
 #include "exec/memory-internal.h"
 #include "exec/ram_addr.h"
+#include "exec/mmu-access-type.h"
+#include "exec/tlb-common.h"
+#include "exec/vaddr.h"
 #include "tcg/tcg.h"
 #include "qemu/error-report.h"
 #include "exec/log.h"
@@ -95,6 +98,54 @@ static inline size_t sizeof_tlb(CPUTLBDescFast *fast)
     return fast->mask + (1 << CPU_TLB_ENTRY_BITS);
 }
 
+static inline uint64_t tlb_read_idx(const CPUTLBEntry *entry,
+                                    MMUAccessType access_type)
+{
+    /* Do not rearrange the CPUTLBEntry structure members. */
+    QEMU_BUILD_BUG_ON(offsetof(CPUTLBEntry, addr_read) !=
+                      MMU_DATA_LOAD * sizeof(uint64_t));
+    QEMU_BUILD_BUG_ON(offsetof(CPUTLBEntry, addr_write) !=
+                      MMU_DATA_STORE * sizeof(uint64_t));
+    QEMU_BUILD_BUG_ON(offsetof(CPUTLBEntry, addr_code) !=
+                      MMU_INST_FETCH * sizeof(uint64_t));
+
+#if TARGET_LONG_BITS == 32
+    /* Use qatomic_read, in case of addr_write; only care about low bits. */
+    const uint32_t *ptr = (uint32_t *)&entry->addr_idx[access_type];
+    ptr += HOST_BIG_ENDIAN;
+    return qatomic_read(ptr);
+#else
+    const uint64_t *ptr = &entry->addr_idx[access_type];
+# if TCG_OVERSIZED_GUEST
+    return *ptr;
+# else
+    /* ofs might correspond to .addr_write, so use qatomic_read */
+    return qatomic_read(ptr);
+# endif
+#endif
+}
+
+static inline uint64_t tlb_addr_write(const CPUTLBEntry *entry)
+{
+    return tlb_read_idx(entry, MMU_DATA_STORE);
+}
+
+/* Find the TLB index corresponding to the mmu_idx + address pair.  */
+static inline uintptr_t tlb_index(CPUState *cpu, uintptr_t mmu_idx,
+                                  vaddr addr)
+{
+    uintptr_t size_mask = cpu->neg.tlb.f[mmu_idx].mask >> CPU_TLB_ENTRY_BITS;
+
+    return (addr >> TARGET_PAGE_BITS) & size_mask;
+}
+
+/* Find the TLB entry corresponding to the mmu_idx + address pair.  */
+static inline CPUTLBEntry *tlb_entry(CPUState *cpu, uintptr_t mmu_idx,
+                                     vaddr addr)
+{
+    return &cpu->neg.tlb.f[mmu_idx].table[tlb_index(cpu, mmu_idx, addr)];
+}
+
 static void tlb_window_reset(CPUTLBDesc *desc, int64_t ns,
                              size_t max_entries)
 {
@@ -1039,7 +1090,7 @@ static inline void tlb_set_dirty1_locked(CPUTLBEntry *tlb_entry,
 
 /* update the TLB corresponding to virtual page vaddr
    so that it is no longer dirty */
-void tlb_set_dirty(CPUState *cpu, vaddr addr)
+static void tlb_set_dirty(CPUState *cpu, vaddr addr)
 {
     int mmu_idx;
 

accel/tcg/ldst_atomicity.c.inc

@@ -9,8 +9,8 @@
  * See the COPYING file in the top-level directory.
  */
 
-#include "host/load-extract-al16-al8.h"
-#include "host/store-insert-al16.h"
+#include "host/load-extract-al16-al8.h.inc"
+#include "host/store-insert-al16.h.inc"
 
 #ifdef CONFIG_ATOMIC64
 # define HAVE_al8          true

accel/tcg/tb-jmp-cache.h

@@ -9,6 +9,9 @@
 #ifndef ACCEL_TCG_TB_JMP_CACHE_H
 #define ACCEL_TCG_TB_JMP_CACHE_H
 
+#include "qemu/rcu.h"
+#include "exec/cpu-common.h"
+
 #define TB_JMP_CACHE_BITS 12
 #define TB_JMP_CACHE_SIZE (1 << TB_JMP_CACHE_BITS)
 

accel/tcg/tcg-accel-ops.c

@@ -37,6 +37,8 @@
 #include "exec/tb-flush.h"
 #include "exec/gdbstub.h"
 
+#include "hw/core/cpu.h"
+
 #include "tcg-accel-ops.h"
 #include "tcg-accel-ops-mttcg.h"
 #include "tcg-accel-ops-rr.h"

accel/tcg/translator.c

@@ -12,6 +12,7 @@
 #include "qemu/error-report.h"
 #include "exec/exec-all.h"
 #include "exec/translator.h"
+#include "exec/cpu_ldst.h"
 #include "exec/plugin-gen.h"
 #include "tcg/tcg-op-common.h"
 #include "internal-target.h"

accel/tcg/user-exec.c

@@ -31,6 +31,7 @@
 #include "tcg/tcg-ldst.h"
 #include "internal-common.h"
 #include "internal-target.h"
+#include "user-retaddr.h"
 
 __thread uintptr_t helper_retaddr;
 

accel/tcg/user-retaddr.h

@@ -0,0 +1,28 @@
+#ifndef ACCEL_TCG_USER_RETADDR_H
+#define ACCEL_TCG_USER_RETADDR_H
+
+#include "qemu/atomic.h"
+
+extern __thread uintptr_t helper_retaddr;
+
+static inline void set_helper_retaddr(uintptr_t ra)
+{
+    helper_retaddr = ra;
+    /*
+     * Ensure that this write is visible to the SIGSEGV handler that
+     * may be invoked due to a subsequent invalid memory operation.
+     */
+    signal_barrier();
+}
+
+static inline void clear_helper_retaddr(void)
+{
+    /*
+     * Ensure that previous memory operations have succeeded before
+     * removing the data visible to the signal handler.
+     */
+    signal_barrier();
+    helper_retaddr = 0;
+}
+
+#endif

bsd-user/freebsd/target_os_elf.h

@@ -22,6 +22,7 @@
 
 #include "target_arch_elf.h"
 #include "elf.h"
+#include "user/tswap-target.h"
 
 #define bsd_get_ncpu() 1 /* until we pull in bsd-proc.[hc] */
 

bsd-user/freebsd/target_os_stack.h

@@ -23,6 +23,7 @@
 #include <sys/param.h>
 #include "target_arch_sigtramp.h"
 #include "qemu/guest-random.h"
+#include "user/tswap-target.h"
 
 /*
  * The initial FreeBSD stack is as follows:

bsd-user/main.c

@@ -36,6 +36,7 @@
 #include "qemu/help_option.h"
 #include "qemu/module.h"
 #include "exec/exec-all.h"
+#include "user/guest-base.h"
 #include "tcg/startup.h"
 #include "qemu/timer.h"
 #include "qemu/envlist.h"

bsd-user/netbsd/target_os_elf.h

@@ -22,6 +22,7 @@
 
 #include "target_arch_elf.h"
 #include "elf.h"
+#include "user/tswap-target.h"
 
 /* this flag is uneffective under linux too, should be deleted */
 #ifndef MAP_DENYWRITE

bsd-user/openbsd/target_os_elf.h

@@ -22,6 +22,7 @@
 
 #include "target_arch_elf.h"
 #include "elf.h"
+#include "user/tswap-target.h"
 
 /* this flag is uneffective under linux too, should be deleted */
 #ifndef MAP_DENYWRITE

bsd-user/signal.c

@@ -21,6 +21,7 @@
 #include "qemu/osdep.h"
 #include "qemu/log.h"
 #include "qemu.h"
+#include "user/tswap-target.h"
 #include "gdbstub/user.h"
 #include "signal-common.h"
 #include "trace.h"

bsd-user/strace.c

@@ -22,6 +22,7 @@
 #include <sys/ioccom.h>
 
 #include "qemu.h"
+#include "user/tswap-target.h"
 
 #include "os-strace.h"  /* OS dependent strace print functions */
 

cpu-target.c

@@ -35,6 +35,7 @@
 #endif
 #include "sysemu/cpus.h"
 #include "sysemu/tcg.h"
+#include "exec/tswap.h"
 #include "exec/replay-core.h"
 #include "exec/cpu-common.h"
 #include "exec/exec-all.h"

disas/disas.c

@@ -6,6 +6,7 @@
 #include "disas/disas.h"
 #include "disas/capstone.h"
 #include "hw/core/cpu.h"
+#include "exec/tswap.h"
 #include "exec/memory.h"
 
 /* Filled in by elfload.c.  Simplistic, but will do for now. */