From: "Daniel P. Berrangé" <berrange@redhat.com>
To: David Hildenbrand <david@redhat.com>
Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>,
Eduardo Habkost <ehabkost@redhat.com>,
"Michael S. Tsirkin" <mst@redhat.com>,
qemu-devel@nongnu.org,
"Dr . David Alan Gilbert" <dgilbert@redhat.com>,
Paolo Bonzini <pbonzini@redhat.com>,
Igor Mammedov <imammedo@redhat.com>,
Marek Kedzierski <mkedzier@redhat.com>
Subject: Re: [PATCH v1 3/3] util/oslib-posix: Support concurrent os_mem_prealloc() invocation
Date: Tue, 20 Jul 2021 15:22:58 +0100 [thread overview]
Message-ID: <YPbcQmgAY+GdsIfb@redhat.com> (raw)
In-Reply-To: <20210714112306.67793-4-david@redhat.com>
On Wed, Jul 14, 2021 at 01:23:06PM +0200, David Hildenbrand wrote:
> Add a mutext to protect the SIGBUS case, as we cannot mess concurrently
typo s/mutext/mutex/
> with the sigbus handler and we have to manage the global variable
> sigbus_memset_context. The MADV_POPULATE_WRITE path can run
> concurrently.
>
> Note that page_mutex and page_cond are shared between concurrent
> invocations, which shouldn't be a problem.
>
> This is a preparation for future virtio-mem prealloc code, which will call
> os_mem_prealloc() asynchronously from an iothread when handling guest
> requests.
Hmm, I'm wondering how the need to temporarily play with SIGBUS
at runtime for mem preallocation will interact with the SIGBUS
handler installed by softmmu/cpus.c.
The SIGBUS handler the preallocation code is installed just
blindly assumes the SIGBUS is related to the preallocation
work being done. This is a fine assumption during initially
startup where we're single threaded and not running guest
CPUs. I'm less clear on whether that's a valid assumption
at runtime once guest CPUs are running.
If the sigbus_handler method in softmmu/cpus.c is doing
something important for QEMU, then why is it ok for us to
periodically disable that handler and replace it with
something else that takes a completely different action ?
Of course with the madvise impl we're bypassing the SIGBUS
dance entirely. This is good for people with new kernels,
but is this SIGBUS stuff safe for older kernels ?
> Signed-off-by: David Hildenbrand <david@redhat.com>
> ---
> util/oslib-posix.c | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> diff --git a/util/oslib-posix.c b/util/oslib-posix.c
> index 60d1da2d6c..181f6bbf1a 100644
> --- a/util/oslib-posix.c
> +++ b/util/oslib-posix.c
> @@ -94,6 +94,7 @@ typedef struct MemsetThread MemsetThread;
>
> /* used by sigbus_handler() */
> static MemsetContext *sigbus_memset_context;
> +static QemuMutex sigbus_mutex;
>
> static QemuMutex page_mutex;
> static QemuCond page_cond;
> @@ -605,12 +606,17 @@ static bool madv_populate_write_possible(char *area, size_t pagesize)
> void os_mem_prealloc(int fd, char *area, size_t memory, int smp_cpus,
> Error **errp)
> {
> + static gsize initialized;
> int ret;
> struct sigaction act, oldact;
> size_t hpagesize = qemu_fd_getpagesize(fd);
> size_t numpages = DIV_ROUND_UP(memory, hpagesize);
> bool use_madv_populate_write;
>
> + if (g_once_init_enter(&initialized)) {
> + qemu_mutex_init(&sigbus_mutex);
> + }
> +
> /*
> * Sense on every invocation, as MADV_POPULATE_WRITE cannot be used for
> * some special mappings, such as mapping /dev/mem.
> @@ -620,6 +626,7 @@ void os_mem_prealloc(int fd, char *area, size_t memory, int smp_cpus,
> }
>
> if (!use_madv_populate_write) {
> + qemu_mutex_lock(&sigbus_mutex);
> memset(&act, 0, sizeof(act));
> act.sa_handler = &sigbus_handler;
> act.sa_flags = 0;
> @@ -646,6 +653,7 @@ void os_mem_prealloc(int fd, char *area, size_t memory, int smp_cpus,
> perror("os_mem_prealloc: failed to reinstall signal handler");
> exit(1);
> }
> + qemu_mutex_unlock(&sigbus_mutex);
> }
> }
>
> --
> 2.31.1
>
>
Regards,
Daniel
--
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org -o- https://fstop138.berrange.com :|
|: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
next prev parent reply other threads:[~2021-07-20 14:24 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-14 11:23 [PATCH v1 0/3] util/oslib-posix: Support MADV_POPULATE_WRITE for os_mem_prealloc() David Hildenbrand
2021-07-14 11:23 ` [PATCH v1 1/3] " David Hildenbrand
2021-07-20 14:08 ` Daniel P. Berrangé
2021-07-20 14:34 ` David Hildenbrand
2021-07-14 11:23 ` [PATCH v1 2/3] util/oslib-posix: Introduce and use MemsetContext for touch_all_pages() David Hildenbrand
2021-07-20 14:27 ` Daniel P. Berrangé
2021-07-14 11:23 ` [PATCH v1 3/3] util/oslib-posix: Support concurrent os_mem_prealloc() invocation David Hildenbrand
2021-07-20 14:22 ` Daniel P. Berrangé [this message]
2021-07-20 14:27 ` David Hildenbrand
2021-07-20 14:31 ` Daniel P. Berrangé
2021-07-20 14:35 ` David Hildenbrand
2021-07-20 13:55 ` [PATCH v1 0/3] util/oslib-posix: Support MADV_POPULATE_WRITE for os_mem_prealloc() Pankaj Gupta
2021-07-20 13:58 ` Pankaj Gupta
2021-07-20 14:45 ` Daniel P. Berrangé
2021-07-21 8:23 ` David Hildenbrand
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=YPbcQmgAY+GdsIfb@redhat.com \
--to=berrange@redhat.com \
--cc=david@redhat.com \
--cc=dgilbert@redhat.com \
--cc=ehabkost@redhat.com \
--cc=imammedo@redhat.com \
--cc=mkedzier@redhat.com \
--cc=mst@redhat.com \
--cc=pankaj.gupta.linux@gmail.com \
--cc=pbonzini@redhat.com \
--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).