dash.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jilles Tjoelker <jilles@stack.nl>
To: Harald van Dijk <harald@gigawatt.nl>
Cc: Vitaly Zuevsky <vitaly.zuevsky@gmail.com>,
	'Andrej Shadura' <andrew@shadura.me>,
	953421@bugs.debian.org, dash@vger.kernel.org,
	'Debian Bug Tracking System' <submit@bugs.debian.org>
Subject: Re: Bug#953421: dash: Resident Set Size growth is unbound (memory leak) on an infinite shell loop
Date: Mon, 30 Mar 2020 00:07:06 +0200	[thread overview]
Message-ID: <20200329220706.GA13241@stack.nl> (raw)
In-Reply-To: <a2efaae8-db1b-39ae-d7c2-8d119a4f14d4@gigawatt.nl>

On Sun, Mar 29, 2020 at 08:06:31PM +0100, Harald van Dijk wrote:
> On 29/03/2020 18:54, Vitaly Zuevsky wrote:
> > I have now fixed this bug locally.

> > The leak is in jobtab array (jobs.c). I concluded that the most
> > logical approach would be eliminating inconsistency between
> > makejob() and dowait() functions. My fix in a forked repo:

> > https://salsa.debian.org/psvz-guest/dash/-/commit/5e3ea90cb3355d1308c482661a471883d36af5e7

> This change is incorrect. The reason dash keeps on allocating memory is
> because dash needs to keep on allocating memory. Consider this script:

>   set -- $(seq 1 100)
>   for i
>   do
>     : &
>     sleep .1
>   done
>   for i
>   do
>     wait %$i
>   done

> This is a valid script and works fine in dash. Your change breaks this by
> not keeping the jobs around long enough, and I hope this test script shows
> that there is no way to keep the jobs around long enough but by allocating
> ever more memory.

I agree that the change is incorrect, but I do not agree that this kind
of script must leak memory. Per POSIX.1-2008 XCU 2.9.3.1 Asynchronous
Lists, an implementation has additional ways to forget about jobs than
just an appropriate completion of the wait utility: if another
asynchronous job is started when $! was not referenced or if the number
of known process IDs would exceed {CHILD_MAX} (which tends to be rather
big, though).

POSIX does not seem to expect using %<jobid> in scripts like this; it
seems highly fragile to me anyway (although $! has problems with process
ID reuse).

FreeBSD sh implements forgetting when $! was not referenced (and the job
has terminated), but not the {CHILD_MAX} limit. This avoids the
increasing memory usage in the example script.

> Your change makes it impossible to keep track of the background process's
> status, but if you do not care about that anyway, you can avoid the
> increasing memory use without modifying dash by launching a background
> process without including it in the current shell's job table, by launching
> it from a subshell:

>   while true
>   do
>     (true &)
>     sleep .1
>   done

Certainly a good idea.

Another option may be to include regular invocations of the wait utility
without parameters, although this is not suitable for all scripts.

-- 
Jilles Tjoelker

  reply	other threads:[~2020-03-29 22:16 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <158376996556.31988.8584094104007124674.reportbug@ec2-34-240-101-198.eu-west-1.compute.amazonaws.com>
     [not found] ` <CACujMDPfs5mJs8CVaxqM6wkCRANYQ71wTUkvHiNvOg+MPSTECQ@mail.gmail.com>
2020-03-29 17:54   ` Bug#953421: dash: Resident Set Size growth is unbound (memory leak) on an infinite shell loop Vitaly Zuevsky
2020-03-29 19:06     ` Harald van Dijk
2020-03-29 22:07       ` Jilles Tjoelker [this message]
2020-03-29 23:07         ` Harald van Dijk
2020-03-31 19:07       ` Vitaly Zuevsky
2020-03-31 21:04         ` Harald van Dijk
2020-04-02 13:18       ` Vitaly Zuevsky

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=20200329220706.GA13241@stack.nl \
    --to=jilles@stack.nl \
    --cc=953421@bugs.debian.org \
    --cc=andrew@shadura.me \
    --cc=dash@vger.kernel.org \
    --cc=harald@gigawatt.nl \
    --cc=submit@bugs.debian.org \
    --cc=vitaly.zuevsky@gmail.com \
    /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).