All of lore.kernel.org
 help / color / mirror / Atom feed
From: Paolo Bonzini <pbonzini@redhat.com>
To: "Michael S. Tsirkin" <mst@redhat.com>, qemu-devel@nongnu.org
Cc: "Peter Maydell" <peter.maydell@linaro.org>,
	"Marc-André Lureau" <mlureau@redhat.com>
Subject: Re: [Qemu-devel] [PULL 2/7] exec: factor out duplicate mmap code
Date: Sun, 25 Oct 2015 16:04:09 +0100	[thread overview]
Message-ID: <562CEF69.4030108@redhat.com> (raw)
In-Reply-To: <1444985258-6605-3-git-send-email-mst@redhat.com>



On 16/10/2015 10:49, Michael S. Tsirkin wrote:
> Anonymous and file-backed RAM allocation are now almost exactly the same.
> 
> Reduce code duplication by moving RAM mmap code out of oslib-posix.c and
> exec.c.
> 
> Reported-by: Marc-André Lureau <mlureau@redhat.com>
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
> Acked-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  include/qemu/mmap-alloc.h | 10 +++++++
>  exec.c                    | 47 +++++++------------------------
>  util/mmap-alloc.c         | 71 +++++++++++++++++++++++++++++++++++++++++++++++
>  util/oslib-posix.c        | 28 +++----------------
>  util/Makefile.objs        |  1 +
>  5 files changed, 96 insertions(+), 61 deletions(-)
>  create mode 100644 include/qemu/mmap-alloc.h
>  create mode 100644 util/mmap-alloc.c
> 
> diff --git a/include/qemu/mmap-alloc.h b/include/qemu/mmap-alloc.h
> new file mode 100644
> index 0000000..56388e6
> --- /dev/null
> +++ b/include/qemu/mmap-alloc.h
> @@ -0,0 +1,10 @@
> +#ifndef QEMU_MMAP_ALLOC
> +#define QEMU_MMAP_ALLOC
> +
> +#include "qemu-common.h"
> +
> +void *qemu_ram_mmap(int fd, size_t size, size_t align, bool shared);
> +
> +void qemu_ram_munmap(void *ptr, size_t size);
> +
> +#endif
> diff --git a/exec.c b/exec.c
> index 7d90a52..4505dc7 100644
> --- a/exec.c
> +++ b/exec.c
> @@ -55,6 +55,9 @@
>  #include "exec/ram_addr.h"
>  
>  #include "qemu/range.h"
> +#ifndef _WIN32
> +#include "qemu/mmap-alloc.h"
> +#endif
>  
>  //#define DEBUG_SUBPAGE
>  
> @@ -84,9 +87,9 @@ static MemoryRegion io_mem_unassigned;
>   */
>  #define RAM_RESIZEABLE (1 << 2)
>  
> -/* An extra page is mapped on top of this RAM.
> +/* RAM is backed by an mmapped file.
>   */
> -#define RAM_EXTRA (1 << 3)
> +#define RAM_FILE (1 << 3)
>  #endif
>  
>  struct CPUTailQ cpus = QTAILQ_HEAD_INITIALIZER(cpus);
> @@ -1188,13 +1191,10 @@ static void *file_ram_alloc(RAMBlock *block,
>      char *filename;
>      char *sanitized_name;
>      char *c;
> -    void *ptr;
> -    void *area = NULL;
> +    void *area;
>      int fd;
>      uint64_t hpagesize;
> -    uint64_t total;
>      Error *local_err = NULL;
> -    size_t offset;
>  
>      hpagesize = gethugepagesize(path, &local_err);
>      if (local_err) {
> @@ -1238,7 +1238,6 @@ static void *file_ram_alloc(RAMBlock *block,
>      g_free(filename);
>  
>      memory = ROUND_UP(memory, hpagesize);
> -    total = memory + hpagesize;
>  
>      /*
>       * ftruncate is not supported by hugetlbfs in older
> @@ -1250,40 +1249,14 @@ static void *file_ram_alloc(RAMBlock *block,
>          perror("ftruncate");
>      }
>  
> -    ptr = mmap(0, total, PROT_NONE, MAP_PRIVATE | MAP_ANONYMOUS,
> -                -1, 0);
> -    if (ptr == MAP_FAILED) {
> -        error_setg_errno(errp, errno,
> -                         "unable to allocate memory range for hugepages");
> -        close(fd);
> -        goto error;
> -    }
> -
> -    offset = QEMU_ALIGN_UP((uintptr_t)ptr, hpagesize) - (uintptr_t)ptr;
> -
> -    area = mmap(ptr + offset, memory, PROT_READ | PROT_WRITE,
> -                (block->flags & RAM_SHARED ? MAP_SHARED : MAP_PRIVATE) |
> -                MAP_FIXED,
> -                fd, 0);
> +    area = qemu_ram_mmap(fd, memory, hpagesize, block->flags & RAM_SHARED);
>      if (area == MAP_FAILED) {
>          error_setg_errno(errp, errno,
>                           "unable to map backing store for hugepages");
> -        munmap(ptr, total);
>          close(fd);
>          goto error;
>      }
>  
> -    if (offset > 0) {
> -        munmap(ptr, offset);
> -    }
> -    ptr += offset;
> -    total -= offset;
> -
> -    if (total > memory + getpagesize()) {
> -        munmap(ptr + memory + getpagesize(),
> -               total - memory - getpagesize());
> -    }
> -
>      if (mem_prealloc) {
>          os_mem_prealloc(fd, area, memory);
>      }
> @@ -1601,7 +1574,7 @@ ram_addr_t qemu_ram_alloc_from_file(ram_addr_t size, MemoryRegion *mr,
>      new_block->used_length = size;
>      new_block->max_length = size;
>      new_block->flags = share ? RAM_SHARED : 0;
> -    new_block->flags |= RAM_EXTRA;
> +    new_block->flags |= RAM_FILE;
>      new_block->host = file_ram_alloc(new_block, size,
>                                       mem_path, errp);
>      if (!new_block->host) {
> @@ -1703,8 +1676,8 @@ static void reclaim_ramblock(RAMBlock *block)
>          xen_invalidate_map_cache_entry(block->host);
>  #ifndef _WIN32
>      } else if (block->fd >= 0) {
> -        if (block->flags & RAM_EXTRA) {
> -            munmap(block->host, block->max_length + getpagesize());
> +        if (block->flags & RAM_FILE) {
> +            qemu_ram_munmap(block->host, block->max_length);
>          } else {
>              munmap(block->host, block->max_length);
>          }
> diff --git a/util/mmap-alloc.c b/util/mmap-alloc.c
> new file mode 100644
> index 0000000..1394269
> --- /dev/null
> +++ b/util/mmap-alloc.c
> @@ -0,0 +1,71 @@
> +/*
> + * Support for RAM backed by mmaped host memory.
> + *
> + * Copyright (c) 2015 Red Hat, Inc.
> + *
> + * Authors:
> + *  Michael S. Tsirkin <mst@redhat.com>
> + *
> + * This work is licensed under the terms of the GNU GPL, version 2 or
> + * later.  See the COPYING file in the top-level directory.
> + */
> +#include <qemu/mmap-alloc.h>
> +#include <sys/types.h>
> +#include <sys/mman.h>
> +#include <assert.h>
> +
> +void *qemu_ram_mmap(int fd, size_t size, size_t align, bool shared)
> +{
> +    /*
> +     * Note: this always allocates at least one extra page of virtual address
> +     * space, even if size is already aligned.
> +     */
> +    size_t total = size + align;
> +    void *ptr = mmap(0, total, PROT_NONE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
> +    size_t offset = QEMU_ALIGN_UP((uintptr_t)ptr, align) - (uintptr_t)ptr;
> +    void *ptr1;
> +
> +    if (ptr == MAP_FAILED) {
> +        return NULL;

Existing code is checking for MAP_FAILED (and NULL is a valid return
value of mmap), so you should return ptr.

> +    }
> +
> +    /* Make sure align is a power of 2 */
> +    assert(!(align & (align - 1)));
> +    /* Always align to host page size */
> +    assert(align >= getpagesize());
> +
> +    ptr1 = mmap(ptr + offset, size, PROT_READ | PROT_WRITE,
> +                MAP_FIXED |
> +                (fd == -1 ? MAP_ANONYMOUS : 0) |
> +                (shared ? MAP_SHARED : MAP_PRIVATE),
> +                fd, 0);
> +    if (ptr1 == MAP_FAILED) {
> +        munmap(ptr, total);
> +        return NULL;

Same here---return ptr1 instead of NULL.

Paolo

> +    }
> +
> +    ptr += offset;
> +    total -= offset;
> +
> +    if (offset > 0) {
> +        munmap(ptr - offset, offset);
> +    }
> +
> +    /*
> +     * Leave a single PROT_NONE page allocated after the RAM block, to serve as
> +     * a guard page guarding against potential buffer overflows.
> +     */
> +    if (total > size + getpagesize()) {
> +        munmap(ptr + size + getpagesize(), total - size - getpagesize());
> +    }
> +
> +    return ptr;
> +}
> +
> +void qemu_ram_munmap(void *ptr, size_t size)
> +{
> +    if (ptr) {
> +        /* Unmap both the RAM block and the guard page */
> +        munmap(ptr, size + getpagesize());
> +    }
> +}
> diff --git a/util/oslib-posix.c b/util/oslib-posix.c
> index a0fcdc2..892d2d8 100644
> --- a/util/oslib-posix.c
> +++ b/util/oslib-posix.c
> @@ -72,6 +72,8 @@ extern int daemon(int, int);
>  #include <sys/sysctl.h>
>  #endif
>  
> +#include <qemu/mmap-alloc.h>
> +
>  int qemu_get_thread_id(void)
>  {
>  #if defined(__linux__)
> @@ -128,10 +130,7 @@ void *qemu_memalign(size_t alignment, size_t size)
>  void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment)
>  {
>      size_t align = QEMU_VMALLOC_ALIGN;
> -    size_t total = size + align;
> -    void *ptr = mmap(0, total, PROT_NONE, MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
> -    size_t offset = QEMU_ALIGN_UP((uintptr_t)ptr, align) - (uintptr_t)ptr;
> -    void *ptr1;
> +    void *ptr = qemu_ram_mmap(-1, size, align, false);
>  
>      if (ptr == MAP_FAILED) {
>          return NULL;
> @@ -141,23 +140,6 @@ void *qemu_anon_ram_alloc(size_t size, uint64_t *alignment)
>          *alignment = align;
>      }
>  
> -    ptr1 = mmap(ptr + offset, size, PROT_READ | PROT_WRITE,
> -                MAP_FIXED | MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
> -    if (ptr1 == MAP_FAILED) {
> -        munmap(ptr, total);
> -        return NULL;
> -    }
> -
> -    ptr += offset;
> -    total -= offset;
> -
> -    if (offset > 0) {
> -        munmap(ptr - offset, offset);
> -    }
> -    if (total > size + getpagesize()) {
> -        munmap(ptr + size + getpagesize(), total - size - getpagesize());
> -    }
> -
>      trace_qemu_anon_ram_alloc(size, ptr);
>      return ptr;
>  }
> @@ -171,9 +153,7 @@ void qemu_vfree(void *ptr)
>  void qemu_anon_ram_free(void *ptr, size_t size)
>  {
>      trace_qemu_anon_ram_free(ptr, size);
> -    if (ptr) {
> -        munmap(ptr, size + getpagesize());
> -    }
> +    qemu_ram_munmap(ptr, size);
>  }
>  
>  void qemu_set_block(int fd)
> diff --git a/util/Makefile.objs b/util/Makefile.objs
> index 114d657..f011b81 100644
> --- a/util/Makefile.objs
> +++ b/util/Makefile.objs
> @@ -1,6 +1,7 @@
>  util-obj-y = osdep.o cutils.o unicode.o qemu-timer-common.o
>  util-obj-$(CONFIG_WIN32) += oslib-win32.o qemu-thread-win32.o event_notifier-win32.o
>  util-obj-$(CONFIG_POSIX) += oslib-posix.o qemu-thread-posix.o event_notifier-posix.o qemu-openpty.o
> +util-obj-$(CONFIG_POSIX) += mmap-alloc.o
>  util-obj-y += envlist.o path.o module.o
>  util-obj-$(call lnot,$(CONFIG_INT128)) += host-utils.o
>  util-obj-y += bitmap.o bitops.o hbitmap.o
> 

  reply	other threads:[~2015-10-25 15:04 UTC|newest]

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-10-16  8:49 [Qemu-devel] [PULL 0/7] vhost, pc features, fixes Michael S. Tsirkin
2015-10-16  8:49 ` [Qemu-devel] [PULL 1/7] intel_iommu: Add support for translation for devices behind bridges Michael S. Tsirkin
2015-10-16  8:49 ` [Qemu-devel] [PULL 2/7] exec: factor out duplicate mmap code Michael S. Tsirkin
2015-10-25 15:04   ` Paolo Bonzini [this message]
2015-10-16  8:49 ` [Qemu-devel] [PULL 3/7] net: don't set native endianness Michael S. Tsirkin
2015-10-16  8:49 ` [Qemu-devel] [PULL 4/7] tests: re-enable vhost-user-test Michael S. Tsirkin
2015-10-16  8:49 ` [Qemu-devel] [PULL 5/7] vhost: add vhost_has_free_slot() interface Michael S. Tsirkin
2015-10-16  8:49 ` [Qemu-devel] [PULL 6/7] pc-dimm: add vhost slots limit check before commiting to hotplug Michael S. Tsirkin
2015-10-16  8:49 ` [Qemu-devel] [PULL 7/7] vhost: fail backend intialization early Michael S. Tsirkin
2015-10-16 18:11 ` [Qemu-devel] [PULL 0/7] vhost, pc features, fixes Peter Maydell

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=562CEF69.4030108@redhat.com \
    --to=pbonzini@redhat.com \
    --cc=mlureau@redhat.com \
    --cc=mst@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-devel@nongnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.