[v2,1/3] accel/tcg: Add CPUState argument to page_unprotect

Commit Message

Richard Henderson April 5, 2025, 3:50 p.m. UTC

In the next patch, page_unprotect will need to pass
the CPUState to tb_invalidate_phys_page_unwind.

Signed-off-by: Richard Henderson <[e-mail]>
---
 include/user/page-protection.h | 2 +-
 accel/tcg/user-exec.c          | 8 +++++---
 linux-user/elfload.c           | 2 +-
 3 files changed, 7 insertions(+), 5 deletions(-)

Comments

Philippe Mathieu-Daudé April 5, 2025, 4:16 p.m. UTC | #1

On 5/4/25 17:50, Richard Henderson wrote:
> In the next patch, page_unprotect will need to pass
> the CPUState to tb_invalidate_phys_page_unwind.
> 
> Signed-off-by: Richard Henderson <[e-mail]>
> ---
>   include/user/page-protection.h | 2 +-
>   accel/tcg/user-exec.c          | 8 +++++---
>   linux-user/elfload.c           | 2 +-
>   3 files changed, 7 insertions(+), 5 deletions(-)

I was not sure this would be acceptable, but this is what we need for
heterogeneous emulation :)

Reviewed-by: Philippe Mathieu-Daudé <[e-mail]>

Patch

diff --git a/include/user/page-protection.h b/include/user/page-protection.h
index d5c8748d49..1de72e31e6 100644
--- a/include/user/page-protection.h
+++ b/include/user/page-protection.h
@@ -16,7 +16,7 @@ 
 #include "exec/target_long.h"
 #include "exec/translation-block.h"
 
-int page_unprotect(tb_page_addr_t address, uintptr_t pc);
+int page_unprotect(CPUState *cpu, tb_page_addr_t address, uintptr_t pc);
 
 int page_get_flags(target_ulong address);
 
diff --git a/accel/tcg/user-exec.c b/accel/tcg/user-exec.c
index 5eef8e7f18..90b345a0cf 100644
--- a/accel/tcg/user-exec.c
+++ b/accel/tcg/user-exec.c
@@ -128,7 +128,7 @@  MMUAccessType adjust_signal_pc(uintptr_t *pc, bool is_write)
 bool handle_sigsegv_accerr_write(CPUState *cpu, sigset_t *old_set,
                                  uintptr_t host_pc, abi_ptr guest_addr)
 {
-    switch (page_unprotect(guest_addr, host_pc)) {
+    switch (page_unprotect(cpu, guest_addr, host_pc)) {
     case 0:
         /*
          * Fault not caused by a page marked unwritable to protect
@@ -584,7 +584,7 @@  bool page_check_range(target_ulong start, target_ulong len, int flags)
                 break;
             }
             /* Asking about writable, but has been protected: undo. */
-            if (!page_unprotect(start, 0)) {
+            if (!page_unprotect(NULL, start, 0)) {
                 ret = false;
                 break;
             }
@@ -704,11 +704,13 @@  void tb_lock_page0(tb_page_addr_t address)
  * immediately exited. (We can only return 2 if the 'pc' argument is
  * non-zero.)
  */
-int page_unprotect(tb_page_addr_t address, uintptr_t pc)
+int page_unprotect(CPUState *cpu, tb_page_addr_t address, uintptr_t pc)
 {
     PageFlagsNode *p;
     bool current_tb_invalidated;
 
+    assert((cpu == NULL) == (pc == 0));
+
     /*
      * Technically this isn't safe inside a signal handler.  However we
      * know this only ever happens in a synchronous SEGV handler, so in
diff --git a/linux-user/elfload.c b/linux-user/elfload.c
index 99811af5e7..7519b6bcda 100644
--- a/linux-user/elfload.c
+++ b/linux-user/elfload.c
@@ -4245,7 +4245,7 @@  static int wmr_page_unprotect_regions(void *opaque, target_ulong start,
         size_t step = MAX(TARGET_PAGE_SIZE, qemu_real_host_page_size());
 
         while (1) {
-            page_unprotect(start, 0);
+            page_unprotect(NULL, start, 0);
             if (end - start <= step) {
                 break;
             }