qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Steve Sistare <steven.sistare@oracle.com>
To: qemu-devel@nongnu.org
Cc: Peter Xu <peterx@redhat.com>, Fabiano Rosas <farosas@suse.de>,
	David Hildenbrand <david@redhat.com>,
	Marcel Apfelbaum <marcel.apfelbaum@gmail.com>,
	Eduardo Habkost <eduardo@habkost.net>,
	Philippe Mathieu-Daude <philmd@linaro.org>,
	Paolo Bonzini <pbonzini@redhat.com>,
	"Daniel P. Berrange" <berrange@redhat.com>,
	Markus Armbruster <armbru@redhat.com>,
	Steve Sistare <steven.sistare@oracle.com>
Subject: [PATCH V4 02/19] physmem: fd-based shared memory
Date: Mon,  2 Dec 2024 05:19:54 -0800	[thread overview]
Message-ID: <1733145611-62315-3-git-send-email-steven.sistare@oracle.com> (raw)
In-Reply-To: <1733145611-62315-1-git-send-email-steven.sistare@oracle.com>

Create MAP_SHARED RAMBlocks by mmap'ing a file descriptor rather than using
MAP_ANON, so the memory can be accessed in another process by passing and
mmap'ing the fd.  This will allow CPR to support memory-backend-ram and
memory-backend-shm objects, provided the user creates them with share=on.

Use memfd_create if available because it has no constraints.  If not, use
POSIX shm_open.  However, this may fail if the shm mount size is too small,
even if the system has free memory, so for backwards compatibility fall
back to qemu_anon_ram_alloc/MAP_ANON on shm_open failure.

For backwards compatibility on Windows, always use MAP_ANON.  share=on has
no purpose there, but the syntax is accepted, and must continue to work.

Exclude Xen.  Xen ignores RAM_SHARED and does its own allocation.

Signed-off-by: Steve Sistare <steven.sistare@oracle.com>
---
 system/physmem.c    | 85 +++++++++++++++++++++++++++++++++++++++++++++++++----
 system/trace-events |  1 +
 2 files changed, 81 insertions(+), 5 deletions(-)

diff --git a/system/physmem.c b/system/physmem.c
index dc1db3a..b0c4b22 100644
--- a/system/physmem.c
+++ b/system/physmem.c
@@ -47,6 +47,7 @@
 #include "qemu/qemu-print.h"
 #include "qemu/log.h"
 #include "qemu/memalign.h"
+#include "qemu/memfd.h"
 #include "exec/memory.h"
 #include "exec/ioport.h"
 #include "sysemu/dma.h"
@@ -2057,6 +2058,70 @@ RAMBlock *qemu_ram_alloc_from_file(ram_addr_t size, MemoryRegion *mr,
 }
 #endif
 
+static bool qemu_memfd_available(void)
+{
+    static int has_memfd = -1;
+
+    if (has_memfd < 0) {
+        has_memfd = qemu_memfd_check(0);
+    }
+    return has_memfd;
+}
+
+/*
+ * We want anonymous shared memory, similar to MAP_SHARED|MAP_ANON, but
+ * some users want the fd.  Allocate shm explicitly to get an fd.
+ */
+static bool qemu_ram_alloc_shared(RAMBlock *new_block, Error **errp)
+{
+    size_t max_length = new_block->max_length;
+    MemoryRegion *mr = new_block->mr;
+    const char *name = memory_region_name(mr);
+    int fd;
+
+    if (qemu_memfd_available()) {
+        fd = qemu_memfd_create(name, max_length + mr->align, 0, 0, 0, errp);
+        if (fd < 0) {
+            return false;
+        }
+    } else if (!qemu_shm_available()) {
+        /*
+         * Backwards compatibility for Windows.  The user may specify a
+         * memory backend with shared=on, and Windows ignores shared.
+         * Fall back to qemu_anon_ram_alloc.
+         */
+        return true;
+    } else {
+        Error *local_err = NULL;
+
+        fd = qemu_shm_alloc(max_length, &local_err);
+        if (fd < 0) {
+            /*
+             * Backwards compatibility in case the shm mount size is too small.
+             * Previous QEMU versions called qemu_anon_ram_alloc for anonymous
+             * shared memory, which could succeed.
+             */
+            error_prepend(&local_err,
+                          "Retrying using MAP_ANON|MAP_SHARED because: ");
+            warn_report_err(local_err);
+            return true;
+        }
+    }
+
+    new_block->mr->align = QEMU_VMALLOC_ALIGN;
+    new_block->host = file_ram_alloc(new_block, max_length, fd, false, 0, errp);
+
+    if (new_block->host) {
+        qemu_set_cloexec(fd);
+        new_block->fd = fd;
+        trace_qemu_ram_alloc_shared(name, max_length, fd, new_block->host);
+        return true;
+    }
+
+    close(fd);
+    return false;
+}
+
 static
 RAMBlock *qemu_ram_alloc_internal(ram_addr_t size, ram_addr_t max_size,
                                   void (*resized)(const char*,
@@ -2089,13 +2154,23 @@ RAMBlock *qemu_ram_alloc_internal(ram_addr_t size, ram_addr_t max_size,
     new_block->page_size = qemu_real_host_page_size();
     new_block->host = host;
     new_block->flags = ram_flags;
+
+    if (!host && !xen_enabled()) {
+        if ((new_block->flags & RAM_SHARED) &&
+            !qemu_ram_alloc_shared(new_block, &local_err)) {
+            goto err;
+        }
+    }
+
     ram_block_add(new_block, &local_err);
-    if (local_err) {
-        g_free(new_block);
-        error_propagate(errp, local_err);
-        return NULL;
+    if (!local_err) {
+        return new_block;
     }
-    return new_block;
+
+err:
+    g_free(new_block);
+    error_propagate(errp, local_err);
+    return NULL;
 }
 
 RAMBlock *qemu_ram_alloc_from_ptr(ram_addr_t size, void *host,
diff --git a/system/trace-events b/system/trace-events
index 5bbc3fb..831a60c 100644
--- a/system/trace-events
+++ b/system/trace-events
@@ -33,6 +33,7 @@ address_space_map(void *as, uint64_t addr, uint64_t len, bool is_write, uint32_t
 find_ram_offset(uint64_t size, uint64_t offset) "size: 0x%" PRIx64 " @ 0x%" PRIx64
 find_ram_offset_loop(uint64_t size, uint64_t candidate, uint64_t offset, uint64_t next, uint64_t mingap) "trying size: 0x%" PRIx64 " @ 0x%" PRIx64 ", offset: 0x%" PRIx64" next: 0x%" PRIx64 " mingap: 0x%" PRIx64
 ram_block_discard_range(const char *rbname, void *hva, size_t length, bool need_madvise, bool need_fallocate, int ret) "%s@%p + 0x%zx: madvise: %d fallocate: %d ret: %d"
+qemu_ram_alloc_shared(const char *name, size_t max_length, int fd, void *host) "%s size %zu fd %d host %p"
 
 # cpus.c
 vm_stop_flush_all(int ret) "ret %d"
-- 
1.8.3.1



  parent reply	other threads:[~2024-12-02 13:43 UTC|newest]

Thread overview: 78+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-12-02 13:19 [PATCH V4 00/19] Live update: cpr-transfer Steve Sistare
2024-12-02 13:19 ` [PATCH V4 01/19] backends/hostmem-shm: factor out allocation of "anonymous shared memory with an fd" Steve Sistare
2024-12-09 17:36   ` Peter Xu
2024-12-12 20:37     ` Steven Sistare
2024-12-02 13:19 ` Steve Sistare [this message]
2024-12-09 19:42   ` [PATCH V4 02/19] physmem: fd-based shared memory Peter Xu
2024-12-12 20:38     ` Steven Sistare
2024-12-12 21:22       ` Peter Xu
2024-12-13 16:41         ` Steven Sistare
2024-12-13 17:05           ` Steven Sistare
2024-12-16 18:19           ` Peter Xu
2024-12-17 21:54             ` Steven Sistare
2024-12-17 22:46               ` Peter Xu
2024-12-18 16:34                 ` Steven Sistare
2024-12-02 13:19 ` [PATCH V4 03/19] memory: add RAM_PRIVATE Steve Sistare
2024-12-09 19:45   ` Peter Xu
2024-12-02 13:19 ` [PATCH V4 04/19] machine: aux-ram-share option Steve Sistare
2024-12-05  8:25   ` Markus Armbruster
2024-12-05 14:24     ` Steven Sistare
2024-12-05 12:08   ` Markus Armbruster
2024-12-05 12:19     ` Markus Armbruster
2024-12-05 14:24       ` Steven Sistare
2024-12-09 19:54   ` Peter Xu
2024-12-12 20:38     ` Steven Sistare
2024-12-12 21:22       ` Peter Xu
2024-12-02 13:19 ` [PATCH V4 05/19] migration: cpr-state Steve Sistare
2024-12-02 13:19 ` [PATCH V4 06/19] physmem: preserve ram blocks for cpr Steve Sistare
2024-12-09 20:07   ` Peter Xu
2024-12-12 20:38     ` Steven Sistare
2024-12-12 22:48       ` Peter Xu
2024-12-13 15:21         ` Peter Xu
2024-12-13 15:30           ` Steven Sistare
2024-12-18 16:34             ` Steven Sistare
2024-12-18 17:00               ` Peter Xu
2024-12-18 20:22                 ` Steven Sistare
2024-12-18 20:33                   ` Peter Xu
2024-12-02 13:19 ` [PATCH V4 07/19] hostmem-memfd: preserve " Steve Sistare
2024-12-18 19:53   ` Steven Sistare
2024-12-18 20:23     ` Peter Xu
2024-12-02 13:20 ` [PATCH V4 08/19] hostmem-shm: " Steve Sistare
2024-12-12 17:38   ` Peter Xu
2024-12-02 13:20 ` [PATCH V4 09/19] migration: incoming channel Steve Sistare
2024-12-05 15:23   ` Markus Armbruster
2024-12-05 20:45     ` Steven Sistare
2024-12-09 12:12       ` Markus Armbruster
2024-12-09 16:36         ` Peter Xu
2024-12-11  9:18         ` Markus Armbruster
2024-12-11 18:58         ` Steven Sistare
2024-12-10 12:46     ` Markus Armbruster
2024-12-02 13:20 ` [PATCH V4 10/19] migration: cpr channel Steve Sistare
2024-12-05 15:37   ` Markus Armbruster
2024-12-05 20:46     ` Steven Sistare
2024-12-06  9:31       ` Markus Armbruster
2024-12-18 19:53         ` Steven Sistare
2024-12-18 20:27           ` Peter Xu
2024-12-18 20:31             ` Steven Sistare
2024-12-02 13:20 ` [PATCH V4 11/19] migration: SCM_RIGHTS for QEMUFile Steve Sistare
2024-12-02 13:20 ` [PATCH V4 12/19] migration: VMSTATE_FD Steve Sistare
2024-12-02 13:20 ` [PATCH V4 13/19] migration: cpr-transfer save and load Steve Sistare
2024-12-02 13:20 ` [PATCH V4 14/19] migration: cpr-transfer mode Steve Sistare
2024-12-04 16:10   ` Steven Sistare
2024-12-10 12:26   ` Markus Armbruster
2024-12-11 22:05     ` Steven Sistare
2024-12-02 13:20 ` [PATCH V4 15/19] tests/migration-test: memory_backend Steve Sistare
2024-12-02 13:20 ` [PATCH V4 16/19] tests/qtest: defer connection Steve Sistare
2024-12-18 21:02   ` Steven Sistare
2024-12-19 15:46   ` Peter Xu
2024-12-19 22:33     ` Steven Sistare
2024-12-02 13:20 ` [PATCH V4 17/19] tests/migration-test: " Steve Sistare
2024-12-02 13:20 ` [PATCH V4 18/19] migration-test: cpr-transfer Steve Sistare
2024-12-18 21:03   ` Steven Sistare
2024-12-19 16:56   ` Peter Xu
2024-12-19 22:34     ` Steven Sistare
2024-12-20 15:41       ` Peter Xu
2024-12-02 13:20 ` [PATCH V4 19/19] migration: cpr-transfer documentation Steve Sistare
2024-12-18 21:03   ` Steven Sistare
2024-12-19 17:02   ` Peter Xu
2024-12-19 22:35     ` Steven Sistare

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=1733145611-62315-3-git-send-email-steven.sistare@oracle.com \
    --to=steven.sistare@oracle.com \
    --cc=armbru@redhat.com \
    --cc=berrange@redhat.com \
    --cc=david@redhat.com \
    --cc=eduardo@habkost.net \
    --cc=farosas@suse.de \
    --cc=marcel.apfelbaum@gmail.com \
    --cc=pbonzini@redhat.com \
    --cc=peterx@redhat.com \
    --cc=philmd@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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).