[Qemu-devel] [RFC v1 04/11] tcg: protect TBContext with tb_lock.

["Alex Bennée" <alex.bennee@linaro.org>]

From: KONRAD Frederic <fred.konrad@greensocs.com>

This protects TBContext with tb_lock to make tb_* thread safe.

We can still have issue with tb_flush in case of multithread TCG:
another CPU can be executing code during a flush.

This can be fixed later by making all other TCG thread exiting before calling
tb_flush().

Signed-off-by: KONRAD Frederic <fred.konrad@greensocs.com>
Message-Id: <1439220437-23957-8-git-send-email-fred.konrad@greensocs.com>
Signed-off-by: Emilio G. Cota <cota@braap.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
[AJB: moved into tree, clean-up history]
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>

---
v7 (FK):
  - Drop a tb_lock in already locked restore_state_to_opc.
v6 (FK):
  - Drop a tb_lock arround tb_find_fast in cpu-exec.c.
---
 cpu-exec.c         |  6 ++++++
 exec.c             |  3 +++
 hw/i386/kvmvapic.c |  3 +++
 translate-all.c    | 32 +++++++++++++++++++++++++-------
 4 files changed, 37 insertions(+), 7 deletions(-)

diff --git a/cpu-exec.c b/cpu-exec.c
index e71113e..3572256 100644
--- a/cpu-exec.c
+++ b/cpu-exec.c
@@ -196,18 +196,24 @@ static void cpu_exec_nocache(CPUState *cpu, int max_cycles,
     if (max_cycles > CF_COUNT_MASK)
         max_cycles = CF_COUNT_MASK;
 
+    tb_lock();
     cpu->tb_invalidated_flag = 0;
     tb = tb_gen_code(cpu, orig_tb->pc, orig_tb->cs_base, orig_tb->flags,
                      max_cycles | CF_NOCACHE
                          | (ignore_icount ? CF_IGNORE_ICOUNT : 0));
     tb->orig_tb = cpu->tb_invalidated_flag ? NULL : orig_tb;
     cpu->current_tb = tb;
+    tb_unlock();
+
     /* execute the generated code */
     trace_exec_tb_nocache(tb, tb->pc);
     cpu_tb_exec(cpu, tb->tc_ptr);
+
+    tb_lock();
     cpu->current_tb = NULL;
     tb_phys_invalidate(tb, -1);
     tb_free(tb);
+    tb_unlock();
 }
 #endif
 
diff --git a/exec.c b/exec.c
index 4f0e5ed..402b751 100644
--- a/exec.c
+++ b/exec.c
@@ -2112,6 +2112,9 @@ static void check_watchpoint(int offset, int len, MemTxAttrs attrs, int flags)
                     continue;
                 }
                 cpu->watchpoint_hit = wp;
+
+                /* Unlocked by cpu_loop_exit or cpu_resume_from_signal.  */
+                tb_lock();
                 tb_check_watchpoint(cpu);
                 if (wp->flags & BP_STOP_BEFORE_ACCESS) {
                     cpu->exception_index = EXCP_DEBUG;
diff --git a/hw/i386/kvmvapic.c b/hw/i386/kvmvapic.c
index c69f374..7c0d542 100644
--- a/hw/i386/kvmvapic.c
+++ b/hw/i386/kvmvapic.c
@@ -14,6 +14,7 @@
 #include "sysemu/kvm.h"
 #include "hw/i386/apic_internal.h"
 #include "hw/sysbus.h"
+#include "tcg/tcg.h"
 
 #define VAPIC_IO_PORT           0x7e
 
@@ -446,6 +447,8 @@ static void patch_instruction(VAPICROMState *s, X86CPU *cpu, target_ulong ip)
     resume_all_vcpus();
 
     if (!kvm_enabled()) {
+        /* Unlocked by cpu_resume_from_signal.  */
+        tb_lock();
         cs->current_tb = NULL;
         tb_gen_code(cs, current_pc, current_cs_base, current_flags, 1);
         cpu_resume_from_signal(cs, NULL);
diff --git a/translate-all.c b/translate-all.c
index 1a02450..1aab243 100644
--- a/translate-all.c
+++ b/translate-all.c
@@ -847,7 +847,9 @@ static void page_flush_tb(void)
 }
 
 /* flush all the translation blocks */
-/* XXX: tb_flush is currently not thread safe */
+/* XXX: tb_flush is currently not thread safe.  System emulation calls it only
+ * with tb_lock taken or from safe_work, so no need to take tb_lock here.
+ */
 void tb_flush(CPUState *cpu)
 {
 #if defined(DEBUG_FLUSH)
@@ -1253,6 +1255,7 @@ void tb_invalidate_phys_page_range(tb_page_addr_t start, tb_page_addr_t end,
     /* we remove all the TBs in the range [start, end[ */
     /* XXX: see if in some cases it could be faster to invalidate all
        the code */
+    tb_lock();
     tb = p->first_tb;
     while (tb != NULL) {
         n = (uintptr_t)tb & 3;
@@ -1320,12 +1323,13 @@ void tb_invalidate_phys_page_range(tb_page_addr_t start, tb_page_addr_t end,
     if (current_tb_modified) {
         /* we generate a block containing just the instruction
            modifying the memory. It will ensure that it cannot modify
-           itself */
+           itself.  cpu_resume_from_signal unlocks tb_lock.  */
         cpu->current_tb = NULL;
         tb_gen_code(cpu, current_pc, current_cs_base, current_flags, 1);
         cpu_resume_from_signal(cpu, NULL);
     }
 #endif
+    tb_unlock();
 }
 
 #ifdef CONFIG_SOFTMMU
@@ -1392,6 +1396,8 @@ static void tb_invalidate_phys_page(tb_page_addr_t addr,
     if (!p) {
         return;
     }
+
+    tb_lock();
     tb = p->first_tb;
 #ifdef TARGET_HAS_PRECISE_SMC
     if (tb && pc != 0) {
@@ -1433,9 +1439,12 @@ static void tb_invalidate_phys_page(tb_page_addr_t addr,
         if (locked) {
             mmap_unlock();
         }
+
+        /* tb_lock released by cpu_resume_from_signal.  */
         cpu_resume_from_signal(cpu, puc);
     }
 #endif
+    tb_unlock();
 }
 #endif
 
@@ -1639,6 +1648,7 @@ void cpu_io_recompile(CPUState *cpu, uintptr_t retaddr)
     target_ulong pc, cs_base;
     uint64_t flags;
 
+    tb_lock();
     tb = tb_find_pc(retaddr);
     if (!tb) {
         cpu_abort(cpu, "cpu_io_recompile: could not find TB for pc=%p",
@@ -1690,11 +1700,15 @@ void cpu_io_recompile(CPUState *cpu, uintptr_t retaddr)
     /* FIXME: In theory this could raise an exception.  In practice
        we have already translated the block once so it's probably ok.  */
     tb_gen_code(cpu, pc, cs_base, flags, cflags);
-    /* TODO: If env->pc != tb->pc (i.e. the faulting instruction was not
-       the first in the TB) then we end up generating a whole new TB and
-       repeating the fault, which is horribly inefficient.
-       Better would be to execute just this insn uncached, or generate a
-       second new TB.  */
+
+    /* This unlocks the tb_lock.
+     *
+     * TODO: If env->pc != tb->pc (i.e. the faulting instruction was not
+     * the first in the TB) then we end up generating a whole new TB and
+     * repeating the fault, which is horribly inefficient.
+     * Better would be to execute just this insn uncached, or generate a
+     * second new TB.
+     */
     cpu_resume_from_signal(cpu, NULL);
 }
 
@@ -1719,6 +1733,8 @@ void dump_exec_info(FILE *f, fprintf_function cpu_fprintf)
     int direct_jmp_count, direct_jmp2_count, cross_page;
     TranslationBlock *tb;
 
+    tb_lock();
+
     target_code_size = 0;
     max_target_code_size = 0;
     cross_page = 0;
@@ -1774,6 +1790,8 @@ void dump_exec_info(FILE *f, fprintf_function cpu_fprintf)
             tcg_ctx.tb_ctx.tb_phys_invalidate_count);
     cpu_fprintf(f, "TLB flush count     %d\n", tlb_flush_count);
     tcg_dump_info(f, cpu_fprintf);
+
+    tb_unlock();
 }
 
 void dump_opcount_info(FILE *f, fprintf_function cpu_fprintf)
-- 
2.7.3