LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
From: Eric Dumazet <dada1@cosmosbay.com>
To: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Cc: Johann Borck <johann.borck@densedata.com>,
	Ulrich Drepper <drepper@redhat.com>,
	Ulrich Drepper <drepper@gmail.com>,
	lkml <linux-kernel@vger.kernel.org>,
	David Miller <davem@davemloft.net>, Andrew Morton <akpm@osdl.org>,
	netdev <netdev@vger.kernel.org>,
	Zach Brown <zach.brown@oracle.com>,
	Christoph Hellwig <hch@infradead.org>,
	Chase Venters <chase.venters@clientec.com>
Subject: Re: [take19 1/4] kevent: Core files.
Date: Tue, 17 Oct 2006 17:32:28 +0200	[thread overview]
Message-ID: <200610171732.28640.dada1@cosmosbay.com> (raw)
In-Reply-To: <20061017150929.GA2288@2ka.mipt.ru>

On Tuesday 17 October 2006 17:09, Evgeniy Polyakov wrote:
> On Tue, Oct 17, 2006 at 04:25:00PM +0200, Eric Dumazet (dada1@cosmosbay.com) 
wrote:
> > On Tuesday 17 October 2006 16:07, Evgeniy Polyakov wrote:
> > > On Tue, Oct 17, 2006 at 03:52:34PM +0200, Eric Dumazet
> > > (dada1@cosmosbay.com)
> >
> > wrote:
> > > > > What about the case, which I described in other e-mail, when in
> > > > > case of the full ring buffer, no new events are written there, and
> > > > > when userspace commits (i.e. marks as ready to be freed or requeued
> > > > > by kernel) some events, new ones will be copied from ready queue
> > > > > into the buffer?
> > > >
> > > > Then, user might receive 'false events', exactly like
> > > > poll()/select()/epoll() can do sometime. IE a 'ready' indication
> > > > while there is no current event available on a particular fd /
> > > > event_source.
> > >
> > > Only if user simultaneously uses oth interfaces and remove even from
> > > the queue when it's copy was in mapped buffer, but in that case it's
> > > user's problem (and if we do want, we can store pointer/index of the
> > > ring buffer entry, so when event is removed from the ready queue (using
> > > kevent_get_events()), appropriate entry in the ring buffer will be
> > > updated to show that it is no longer valid.
> > >
> > > > This should be safe, since those programs already ignore read()
> > > > returns -EAGAIN and other similar things.
> > > >
> > > > Programmer prefers to receive two 'event available' indications than
> > > > ZERO (and be stuck for infinite time). Of course, hot path (normal
> > > > cases) should return one 'event' only.
> > > >
> > > > In order words, being ultra fast 99.99 % of the time, but being able
> > > > to block forever once in a while is not an option.
> > >
> > > Have I missed something? It looks like the only problematic situation
> > > is described above when user simultaneously uses both interfaces.
> >
> > In my point of view, user of the 'mmaped ring buffer' should be prepared
> > to use both interfaces. Or else you are forced to presize the ring buffer
> > to insane limits.
> >
> > That is :
> > - Most of the time, we expect consuming events via mmaped ring buffer and
> > no syscalls.
> > - In case we notice a 'mmaped ring buffer overflow', syscalls to
> > get/consume events that could not be stored in mmaped buffer (but queued
> > by kevent subsystem). If not stored by kevent subsystem (memory failure
> > ?), revert to poll() to fetch all 'missed fds' in one row. Go back to
> > normal mode.
>
> kevent uses smaller amount of memory than epoll() per event, so it is very
> unlikely that it will be impossible to store new event there and epoll()
> will succeed. The same can be applied to poll(), which allocates the
> whole table in syscall.
>
> > - In case of empty ring buffer (or no mmap support at all, because this
> > app doesnt expect lot of events per time unit, or because kevent dont
> > have mmap support) : Be able to syscall and wait for an event.
>
> So the most complex case is when user is going to use both interfaces,
> and it's steps when mapped ring buffer has overflow.
> In that case user can either read and mark some events as ready in ring
> buffer (the latter is being done through special syscall), so kevent
> core will put there new ready events.
> User can also get events using usual syscall, in that case events in
> ring buffer must be updated - and actually I implemented mapped buffer
> in the way which allows to remove events from the queue - queue is a
> FIFO, and the first entry to be obtained through syscall is _always_ the
> first entry in the ring buffer.
>
> So when user reads event through syscall (no matter if we are in overflow
> case or not), even being read is easily accessible in the ring buffer.
>
> So I propose following design for ring buffer (quite simple):
> kernelspace maintains two indexes - to the first and the last events in
> the ring buffer (and maximum size of the buffer of course).
> When new event is marked as ready, some info is being copied into ring
> buffer and index of the last entry is increased.
> When event is being read through syscall it is _guaranteed_ that that
> event will be at the position pointed by the index of the first
> element, that index is then increased (thus opening new slot in the
> buffer).
> If index of the last entry reaches (with possible wrapping) index of the
> first entry, that means that overflow has happend. In this case no new
> events can be copied into ring buffer, so they are only placed into
> ready queue (accessible through syscall kevent_get_events()).
>
> When user calls kevent_get_events() it will obtain the first element
> (pointed by index of the first element in the ring buffer), and if there
> is ready event, which is not placed into the ring buffer, it is
> copied (with appropriate update of the last index and new overflow
> condition).

Well, I'm not sure its good to do this 'move one event from ready list to slot 
X', one by one, because this event will likely be flushed out of processor 
cache (because we will have to consume 4096 events before reaching this one). 
I think its better to batch this kind of 'push XX events' later, XX being 
small enough not to waste CPU cache, and when ring buffer is empty again.

mmap buffer is good for latency and minimum synchro between user thread and 
kernel producer. But once we hit an 'overflow', it is better to revert to a 
mode feeding XX events per syscall, to be sure it fits CPU caches : The user 
thread will do the copy between kernel memory to user memory, and this thread 
will shortly use those events in user land.

BTW, maintaining coherency on mmap buffer is expensive : once a event is 
copied to mmap buffer, kernel has to issue a smp_mb() before updating the 
index, so that a user thread wont start to consume an event with random 
values because its CPU see the update on index before updates on data.

Once all the queue is flushed in efficient way, we can switch to mmap mode 
again.

Eric

  reply	other threads:[~2006-10-17 15:32 UTC|newest]

Thread overview: 68+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <115a6230591036@2ka.mipt.ru>
2006-09-12  8:41 ` [take18 0/4] kevent: Generic event handling mechanism Evgeniy Polyakov
2006-09-12  8:41   ` [take18 1/4] kevent: Core files Evgeniy Polyakov
2006-09-12  8:41     ` [take18 2/4] kevent: poll/select() notifications Evgeniy Polyakov
2006-09-12  8:41       ` [take18 3/4] kevent: Socket notifications Evgeniy Polyakov
2006-09-12  8:41         ` [take18 4/4] kevent: Timer notifications Evgeniy Polyakov
2006-09-20  9:35 ` [take19 0/4] kevent: Generic event handling mechanism Evgeniy Polyakov
2006-09-20  9:35   ` [take19 1/4] kevent: Core files Evgeniy Polyakov
2006-09-20  9:35     ` [take19 2/4] kevent: poll/select() notifications Evgeniy Polyakov
2006-09-20  9:35       ` [take19 3/4] kevent: Socket notifications Evgeniy Polyakov
2006-09-20  9:35         ` [take19 4/4] kevent: Timer notifications Evgeniy Polyakov
2006-10-04  6:34     ` [take19 1/4] kevent: Core files Ulrich Drepper
2006-10-04  6:48       ` Evgeniy Polyakov
2006-10-04 17:57         ` Ulrich Drepper
2006-10-05  8:57           ` Evgeniy Polyakov
2006-10-05  9:56             ` Eric Dumazet
2006-10-05 10:21               ` Evgeniy Polyakov
2006-10-05 10:45                 ` Eric Dumazet
2006-10-05 10:55                   ` Evgeniy Polyakov
2006-10-05 12:09                     ` Eric Dumazet
2006-10-05 12:37                       ` Evgeniy Polyakov
2006-10-15 23:22                         ` Ulrich Drepper
2006-10-16  7:33                           ` Evgeniy Polyakov
2006-10-16 10:16                             ` Ulrich Drepper
2006-10-16 11:23                               ` Evgeniy Polyakov
2006-10-17  5:10                           ` Johann Borck
2006-10-17  5:59                             ` Chase Venters
2006-10-17 10:42                               ` Evgeniy Polyakov
2006-10-17 13:12                                 ` Chase Venters
2006-10-17 13:35                                   ` Evgeniy Polyakov
2006-10-17 10:39                             ` Evgeniy Polyakov
2006-10-17 13:19                               ` Eric Dumazet
2006-10-17 13:42                                 ` Evgeniy Polyakov
2006-10-17 13:52                                   ` Eric Dumazet
2006-10-17 14:07                                     ` Evgeniy Polyakov
2006-10-17 14:25                                       ` Eric Dumazet
2006-10-17 15:09                                         ` Evgeniy Polyakov
2006-10-17 15:32                                           ` Eric Dumazet [this message]
2006-10-17 16:01                                             ` Evgeniy Polyakov
2006-10-17 16:26                                               ` Eric Dumazet
2006-10-17 16:35                                                 ` Evgeniy Polyakov
2006-10-17 16:45                                                   ` Eric Dumazet
2006-10-18  4:10                                                     ` Evgeniy Polyakov
2006-10-18  4:45                                                       ` Eric Dumazet
2006-10-17 15:33                                         ` Hans Henrik Happe
2006-10-05 14:01                 ` Hans Henrik Happe
2006-10-05 14:15                   ` Evgeniy Polyakov
2006-10-05 15:07                     ` Hans Henrik Happe
2006-09-22 19:22   ` [take19 0/4] kevent: Generic event handling mechanism Andrew Morton
2006-09-23  4:23     ` Evgeniy Polyakov
2006-10-04  6:09       ` Ulrich Drepper
2006-10-04  6:10         ` Ulrich Drepper
2006-10-04  6:27           ` Evgeniy Polyakov
2006-10-04  6:24         ` Evgeniy Polyakov
2006-09-26 15:54     ` Christoph Hellwig
2006-09-27  4:46       ` Evgeniy Polyakov
2006-09-27 15:09   ` Evgeniy Polyakov
2006-10-04  4:50     ` Ulrich Drepper
2006-10-04  4:55       ` Evgeniy Polyakov
2006-10-04  7:33         ` Ulrich Drepper
2006-10-04  7:48           ` Evgeniy Polyakov
2006-10-04 17:20             ` Ulrich Drepper
2006-10-05  9:02               ` Evgeniy Polyakov
2006-10-05 14:45                 ` Ulrich Drepper
2006-10-06  8:36                   ` Evgeniy Polyakov
2006-10-15 22:43                     ` Ulrich Drepper
2006-10-16  7:23                       ` Evgeniy Polyakov
2006-10-16  9:59                         ` Ulrich Drepper
2006-10-16 10:38                           ` Evgeniy Polyakov

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=200610171732.28640.dada1@cosmosbay.com \
    --to=dada1@cosmosbay.com \
    --cc=akpm@osdl.org \
    --cc=chase.venters@clientec.com \
    --cc=davem@davemloft.net \
    --cc=drepper@gmail.com \
    --cc=drepper@redhat.com \
    --cc=hch@infradead.org \
    --cc=johann.borck@densedata.com \
    --cc=johnpol@2ka.mipt.ru \
    --cc=linux-kernel@vger.kernel.org \
    --cc=netdev@vger.kernel.org \
    --cc=zach.brown@oracle.com \
    --subject='Re: [take19 1/4] kevent: Core files.' \
    /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

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).