Re: Hard-limits for kern.maxproc

30 Jan 2009

      site_archiver@lists.apple.com
Delivered-To: darwin-kernel@lists.apple.com
Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:received:in-reply-to:references :date:message-id:subject:from:to:cc:content-type :content-transfer-encoding; bh=h/9rf/u2dB7AG2wQhguufgl0KJDeDsfYS5uyuSyVaW4=; b=A6xfycBqlzI0b/ftLrVubpSxlpsHaSrk4lyScZ5bfEKEufOYwflgbYob/ZpUpdGgc8 7Bs7FXoyl6ckP69c/uMvrxKlBHgiJTJJhVSD3Wel2iHZeeo2N3TqgzuWYJEqtRoBDfy0 bur+NZL+X8eMqxCpabTCM0qwerSDjbsXDaSaE=
Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type:content-transfer-encoding; b=N0TjT3xtUvbwy54btZ8irAJlXhnABdWjPhPWq0AEAE6fu9qwkJLUwrzhvw6vyT7+Aa SKCUMj7AHYDpp8Kk6MCXFAcpYcP5hislLx7o/uNQJyNnGhX1dURWmWou7ntEH0pEX3kF Rrh3V+3RonCg9cqGLUgWuYxUpdwIwmoCgB1aQ=

Thanks Terry for your answer, on the iPhone you should think to do a
contest personally I give up after 50 words :)

anyway you gave two solutions for the kernel concern:

1- rebuilding with setting up hard_limit
2- write a tool with lib kvm

as I found the second point a good TP (Travaux Pratiques) to do
I started to play in this way:

Materials & Methods
leopard-x86, boot-args="kmem=1", Libkvm-25(10.4.11)

it's working but not in an easy way indeed hard_maxproc is defined as
__private_extern__ so hidden

kvm_t *kd;

struct nlist nl[] = {
	{ "_maxproc" },
	{ "" }
};

int main(int argc, char *argv[])
{
	int _maxproc;
	unsigned long offset;

	if (NULL == (kd = kvm_open("/mach_kernel", NULL, NULL, O_RDWR, "kvm_open"))) {
		return 1;
	}

	if (-1 == kvm_nlist(kd, nl)) {
		warn("kvm_nlist");
		return 1;
	}

	if ((offset =  nl[0].n_value)) {
		if (kvm_read(kd, offset, &_maxproc, sizeof _maxproc) < 0) {
			warn("kvm_read");
			return 1;
		}

		printf("%i \n" , _maxproc);
	}

	return 0;
}

, and anyway this test will be apply

if (count > hard_maxproc)
  count = hard_maxproc;

so the only way I found it's to calculate the syscall addr and overwrite it

if (kvm_write(kd, addr + offset,

the result is really ugly and hackish, so the cleanest way is to
rebuild the kernel with an hard_maxproc value, maybe I missed a point
here

conclusion, even if mail clients are a bit stupid with IMAP protocol,
I think there is still something to do with the server conf
maybe in tracking dead processes and closing them.

Cheers!

On Fri, Jan 30, 2009 at 12:59 PM, Terry Lambert <tlambert@apple.com> wrote:
...
On Jan 30, 2009, at 5:29 AM, Lassi A. Tuura wrote:
...
...
If Mail does this  then you *definitely* should file a bug. He stated
here, and in his numerous posts on the other mailing lists, that he had only
a few Mac OS X client machines, so I had assumed he was thinking about some
other client.
Mail.app definitely spawns large numbers of TCP connections to the IMAP
server (dovecot 1.x) for me.  On "Go online" I see one IMAP connection per
folder (*).  Longer term Mail.app appears to keep 5-15 IMAP connections open
at any one time.  My use appears to generate, from daily stats, on average
one connection every 5-10 seconds or so.
Again, not the list...
Speaking purely as someone who has implemented mail clients before, I can
easily see how certain types of object oriented encapsulation of transport
sessions within mailbox sessions as instances rather than references could
result in a mailbox:transport cardinality of 1:1 instead of N:1, as a design
simplification of the object model used within the mail client.
Speaking purely as a former contributor to the Cyrus IMAP server project and
former technical lead on several commercial IMAP server products, I think
you should file a bug.
Can we be done now?
-- Terry
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-kernel mailing list      (Darwin-kernel@lists.apple.com)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/darwin-kernel/openspecies%40gmail.com
This email sent to openspecies@gmail.com
--
-mmw
 _______________________________________________
Do not post admin requests to the list. They will be ignored.
Darwin-kernel mailing list      (Darwin-kernel@lists.apple.com)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.a...

This email sent to site_archiver@lists.apple.com

mm w

tags

participants (1)