Skip to contentSkip to conversation

bsd-user: Baby Steps towards eliminating qemu_host_page_size, et al

PATCH In review Warner Losh 2 participants 5 messages Jun 7, 2024

lore
Share & tools
lore
#0 lore raw

cover bsd-user: Baby Steps towards eliminating qemu_host_page_size, et al

First baby-steps towards eliminating qemu_host_page_size: tackle the reserve_va calculation (which is easier to copy from linux-user than to fix).

Warner Losh (3):
linux-user: Adjust comment to reflect the code.
bsd-user: port linux-user:ff8a8bbc2ad1 for variable page sizes
bsd-user: Catch up to run-time reserved_va math

 bsd-user/main.c   | 51 ++++++++++++++++++++++++++++++++++++-----------
 linux-user/main.c |  8 ++++----
 2 files changed, 43 insertions(+), 16 deletions(-)
Signature · 1 line
2.43.0
#1 lore raw

PATCH 1/3 linux-user: Adjust comment to reflect the code.

If the user didn't specify a reserved_va, there's an else for 64-bit host 32-bit (or fewer) target to reserve 32-bits of address space. Update the comments to reflect this, and rejustify comment to 80 columns.

---
 linux-user/main.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
linux-user/main.c +4 -4
index 94e4c47f052..94c99a1366f 100644
@@ -814,10 +814,10 @@ int main(int argc, char **argv, char **envp)
     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.
+     * 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. Unless we're running a
+     * target address space of 32 or fewer bits on a host with 64 bits.
      */
     max_reserved_va = MAX_RESERVED_VA(cpu);
     if (reserved_va != 0) {
Signature · 1 line
2.43.0
#2 lore raw

PATCH 2/3 bsd-user: port linux-user:ff8a8bbc2ad1 for variable page sizes

In reply to Warner Losh

Bring in Richard Henderson's ff8a8bbc2ad1 to finalize the page size to allow TARGET_PAGE_BITS_VARY. bsd-user's "blitz" fork has aarch64 support, which is now variable page size. Add support for it here, even though it's effectively a nop in upstream qemu.

---
 bsd-user/main.c | 12 ++++++++++++
 1 file changed, 12 insertions(+)
bsd-user/main.c +12 -0
index 29a629d8779..d685734d087 100644
@@ -46,6 +46,7 @@
 #include "crypto/init.h"
 #include "qemu/guest-random.h"
 #include "gdbstub/user.h"
+#include "exec/page-vary.h"
 
 #include "host-os.h"
 #include "target_arch_cpu.h"
@@ -291,6 +292,7 @@ int main(int argc, char **argv)
     char **target_environ, **wrk;
     envlist_t *envlist = NULL;
     char *argv0 = NULL;
+    int host_page_size;
 
     adjust_ssize();
 
@@ -476,6 +478,16 @@ int main(int argc, char **argv)
                                  opt_one_insn_per_tb, &error_abort);
         ac->init_machine(NULL);
     }
+
+    /*
+     * Finalize page size before creating CPUs.
+     * This will do nothing if !TARGET_PAGE_BITS_VARY.
+     * The most efficient setting is to match the host.
+     */
+    host_page_size = qemu_real_host_page_size();
+    set_preferred_target_page_bits(ctz32(host_page_size));
+    finalize_target_page_bits();
+
     cpu = cpu_create(cpu_type);
     env = cpu_env(cpu);
     cpu_reset(cpu);
Signature · 1 line
2.43.0
#3 lore raw

PATCH 3/3 bsd-user: Catch up to run-time reserved_va math

In reply to Warner Losh

Catch up to linux-user's 8f67b9c694d0, 13c13397556a, 2f7828b57293, and 95059f9c313a by Richard Henderson which made reserved_va a run-time calculation, defaulting to nothing except in the case of 64-bit host 32-bit target. Also include the adjustment of the comment heading that work submitted in the same patch stream. Since this is a direct copy, squash it into one patch rather than follow the Linux evolution since breaking this down further at this point doesn't make sense for this "new code".

---
 bsd-user/main.c | 39 +++++++++++++++++++++++++++------------
 1 file changed, 27 insertions(+), 12 deletions(-)
bsd-user/main.c +27 -12
index d685734d087..dcad266c2c9 100644
@@ -77,25 +77,16 @@ bool have_guest_base;
 # if HOST_LONG_BITS > TARGET_VIRT_ADDR_SPACE_BITS
 #  if TARGET_VIRT_ADDR_SPACE_BITS == 32 && \
       (TARGET_LONG_BITS == 32 || defined(TARGET_ABI32))
-#   define MAX_RESERVED_VA  0xfffffffful
+#   define MAX_RESERVED_VA(CPU)  0xfffffffful
 #  else
-#   define MAX_RESERVED_VA  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
+#   define MAX_RESERVED_VA(CPU)  ((1ul << TARGET_VIRT_ADDR_SPACE_BITS) - 1)
 #  endif
 # else
-#  define MAX_RESERVED_VA  0
+#  define MAX_RESERVED_VA(CPU)  0
 # endif
 #endif
 
-/*
- * That said, 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
-unsigned long reserved_va = MAX_RESERVED_VA;
-#else
 unsigned long reserved_va;
-#endif
 
 const char *interp_prefix = CONFIG_QEMU_INTERP_PREFIX;
 const char *qemu_uname_release;
@@ -293,6 +284,7 @@ int main(int argc, char **argv)
     envlist_t *envlist = NULL;
     char *argv0 = NULL;
     int host_page_size;
+    unsigned long max_reserved_va;
 
     adjust_ssize();
 
@@ -493,6 +485,29 @@ int main(int argc, char **argv)
     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. Unless we're running a
+     * target address space of 32 or fewer bits on a host with 64 bits.
+     */
+    max_reserved_va = MAX_RESERVED_VA(cpu);
+    if (reserved_va != 0) {
+        if ((reserved_va + 1) % host_page_size) {
+            char *s = size_to_str(host_page_size);
+            fprintf(stderr, "Reserved virtual address not aligned mod %s\n", s);
+            g_free(s);
+            exit(EXIT_FAILURE);
+        }
+        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) {
+        /* MAX_RESERVED_VA + 1 is a large power of 2, so is aligned. */
+        reserved_va = max_reserved_va;
+    }
+
     if (getenv("QEMU_STRACE")) {
         do_strace = 1;
     }
Signature · 1 line
2.43.0