[Qemu-devel] [PATCH 0/2] linux-user/arm: Adjust MAX_RESERVED_VA for M-profile

[Qemu-devel] [PATCH 0/2] linux-user/arm: Adjust MAX_RESERVED_VA for M-profile

[Richard Henderson]

This is inspired by the discussion in

   https://bugs.launchpad.net/qemu/+bug/1840922

Previously I suggested a new CPUClass hook, but when I went
to implement that seemed like overkill.


r~


Richard Henderson (2):
  linux-user: Pass CPUState to MAX_RESERVED_VA
  linux-user/arm: Adjust MAX_RESERVED_VA for M-profile

 linux-user/arm/target_cpu.h | 24 ++++++++++++++++++---
 linux-user/main.c           | 43 +++++++++++++++++++++----------------
 2 files changed, 46 insertions(+), 21 deletions(-)

-- 
2.17.1

[Qemu-devel] [PATCH 1/2] linux-user: Pass CPUState to MAX_RESERVED_VA

[Richard Henderson]

Turn the scalar macro into a functional macro.  Move the creation
of the cpu up a bit within main() so that we can pass it to the
invocation of MAX_RESERVED_VA.  Delay the validation of the -R
parameter until MAX_RESERVED_VA is computed.

So far no changes to any of the MAX_RESERVED_VA macros to actually
use the cpu in any way, but ARM will need it.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 linux-user/arm/target_cpu.h |  2 +-
 linux-user/main.c           | 43 +++++++++++++++++++++----------------
 2 files changed, 26 insertions(+), 19 deletions(-)

diff --git a/linux-user/arm/target_cpu.h b/linux-user/arm/target_cpu.h
index 8a3764919a..279ea532d5 100644
--- a/linux-user/arm/target_cpu.h
+++ b/linux-user/arm/target_cpu.h
@@ -21,7 +21,7 @@
 
 /* We need to be able to map the commpage.
    See validate_guest_space in linux-user/elfload.c.  */
-#define MAX_RESERVED_VA  0xffff0000ul
+#define MAX_RESERVED_VA(CPU)  0xffff0000ul
 
 static inline void cpu_clone_regs(CPUARMState *env, target_ulong newsp)
 {
diff --git a/linux-user/main.c b/linux-user/main.c
index 47917bbb20..35da3bf14c 100644
--- a/linux-user/main.c
+++ b/linux-user/main.c
@@ -78,12 +78,12 @@ int have_guest_base;
       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
 /* There are a number of places where we assign reserved_va to a variable
    of type abi_ulong and expect it to fit.  Avoid the last page.  */
-#   define MAX_RESERVED_VA  (0xfffffffful & TARGET_PAGE_MASK)
+#   define MAX_RESERVED_VA(CPU)  (0xfffffffful & TARGET_PAGE_MASK)
 #  else
-#   define MAX_RESERVED_VA  (1ul << TARGET_VIRT_ADDR_SPACE_BITS)
+#   define MAX_RESERVED_VA(CPU)  (1ul << TARGET_VIRT_ADDR_SPACE_BITS)
 #  endif
 # else
-#  define MAX_RESERVED_VA  0
+#  define MAX_RESERVED_VA(CPU)  0
 # endif
 #endif
 
@@ -357,8 +357,7 @@ static void handle_arg_reserved_va(const char *arg)
         unsigned long unshifted = reserved_va;
         p++;
         reserved_va <<= shift;
-        if (reserved_va >> shift != unshifted
-            || (MAX_RESERVED_VA && reserved_va > MAX_RESERVED_VA)) {
+        if (reserved_va >> shift != unshifted) {
             fprintf(stderr, "Reserved virtual address too big\n");
             exit(EXIT_FAILURE);
         }
@@ -607,6 +606,7 @@ int main(int argc, char **argv, char **envp)
     int i;
     int ret;
     int execfd;
+    unsigned long max_reserved_va;
 
     error_init(argv[0]);
     module_call_init(MODULE_INIT_TRACE);
@@ -672,24 +672,31 @@ int main(int argc, char **argv, char **envp)
     /* init tcg before creating CPUs and to get qemu_host_page_size */
     tcg_exec_init(0);
 
-    /* Reserving *too* much vm space via mmap can run into problems
-       with rlimits, oom due to page table creation, etc.  We will still try it,
-       if directed by the command-line option, but not by default.  */
-    if (HOST_LONG_BITS == 64 &&
-        TARGET_VIRT_ADDR_SPACE_BITS <= 32 &&
-        reserved_va == 0) {
-        /* reserved_va must be aligned with the host page size
-         * as it is used with mmap()
-         */
-        reserved_va = MAX_RESERVED_VA & qemu_host_page_mask;
-    }
-
     cpu = cpu_create(cpu_type);
     env = cpu->env_ptr;
     cpu_reset(cpu);
-
     thread_cpu = cpu;
 
+    /*
+     * Reserving too much vm space via mmap can run into problems
+     * with rlimits, oom due to page table creation, etc.  We will
+     * still try it, if directed by the command-line option, but
+     * not by default.
+     */
+    max_reserved_va = MAX_RESERVED_VA(cpu);
+    if (reserved_va != 0) {
+        if (max_reserved_va && reserved_va > max_reserved_va) {
+            fprintf(stderr, "Reserved virtual address too big\n");
+            exit(EXIT_FAILURE);
+        }
+    } else if (HOST_LONG_BITS == 64 && TARGET_VIRT_ADDR_SPACE_BITS <= 32) {
+        /*
+         * reserved_va must be aligned with the host page size
+         * as it is used with mmap()
+         */
+        reserved_va = max_reserved_va & qemu_host_page_mask;
+    }
+
     if (getenv("QEMU_STRACE")) {
         do_strace = 1;
     }
-- 
2.17.1

[Qemu-devel] [PATCH 2/2] linux-user/arm: Adjust MAX_RESERVED_VA for M-profile

[Richard Henderson]

Limit the virtual address space for M-profile cpus to 2GB,
so that we avoid all of the magic addresses in the top half
of the M-profile system map.

Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
 linux-user/arm/target_cpu.h | 24 +++++++++++++++++++++---
 1 file changed, 21 insertions(+), 3 deletions(-)

diff --git a/linux-user/arm/target_cpu.h b/linux-user/arm/target_cpu.h
index 279ea532d5..3f79356a07 100644
--- a/linux-user/arm/target_cpu.h
+++ b/linux-user/arm/target_cpu.h
@@ -19,9 +19,27 @@
 #ifndef ARM_TARGET_CPU_H
 #define ARM_TARGET_CPU_H
 
-/* We need to be able to map the commpage.
-   See validate_guest_space in linux-user/elfload.c.  */
-#define MAX_RESERVED_VA(CPU)  0xffff0000ul
+static inline unsigned long arm_max_reserved_va(CPUState *cs)
+{
+    ARMCPU *cpu = ARM_CPU(cs);
+
+    if (arm_feature(&cpu->env, ARM_FEATURE_M)) {
+        /*
+         * There are magic return addresses above 0xfe000000,
+         * and in general a lot of M-profile system stuff in
+         * the high addresses.  Restrict linux-user to the
+         * cached write-back RAM in the system map.
+         */
+        return 0x80000000ul;
+    } else {
+        /*
+         * We need to be able to map the commpage.
+         * See validate_guest_space in linux-user/elfload.c.
+         */
+        return 0xffff0000ul;
+    }
+}
+#define MAX_RESERVED_VA  arm_max_reserved_va
 
 static inline void cpu_clone_regs(CPUARMState *env, target_ulong newsp)
 {
-- 
2.17.1

Re: [Qemu-devel] [PATCH 0/2] linux-user/arm: Adjust MAX_RESERVED_VA for M-profile

[Peter Maydell]

On Thu, 22 Aug 2019 at 19:59, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> This is inspired by the discussion in
>
>    https://bugs.launchpad.net/qemu/+bug/1840922
>
> Previously I suggested a new CPUClass hook, but when I went
> to implement that seemed like overkill.
>
>
> r~
>
>
> Richard Henderson (2):
>   linux-user: Pass CPUState to MAX_RESERVED_VA
>   linux-user/arm: Adjust MAX_RESERVED_VA for M-profile
>

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>

thanks
-- PMM

Re: [Qemu-devel] [PATCH 0/2] linux-user/arm: Adjust MAX_RESERVED_VA for M-profile

[Laurent Vivier]

Le 03/09/2019 à 15:35, Peter Maydell a écrit :
> On Thu, 22 Aug 2019 at 19:59, Richard Henderson
> <richard.henderson@linaro.org> wrote:
>>
>> This is inspired by the discussion in
>>
>>    https://bugs.launchpad.net/qemu/+bug/1840922
>>
>> Previously I suggested a new CPUClass hook, but when I went
>> to implement that seemed like overkill.
>>
>>
>> r~
>>
>>
>> Richard Henderson (2):
>>   linux-user: Pass CPUState to MAX_RESERVED_VA
>>   linux-user/arm: Adjust MAX_RESERVED_VA for M-profile
>>
> 
> Reviewed-by: Peter Maydell <peter.maydell@linaro.org>

Both applied to my linux-user branch.

Thanks,
Laurent