March 2012 - Darwin-kernel - lists.apple.com

format of sysctl( { CTL_KERN, KERN_PROCARGS, pid } )
by Gerriet M. Denkmann 31 Mar '12

31 Mar '12

Is the output format of format of sysctl( { CTL_KERN, KERN_PROCARGS, pid } ) somewhere documented? The output seems to be mostly Utf8-strings, with some ints thrown in. These ints seem to have their lower 16 bits = zero. Kind regards, Gerriet. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Kauth event for write operation
by simran charley 30 Mar '12

30 Mar '12

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

What the swap??
by Gerriet M. Denkmann 29 Mar '12

29 Mar '12

I just did: sum = 0 for all processes (as reported by sysctl) (without 0 kernel_task) do: for all memory regions do: vm_region_64( VM_REGION_EXTENDED_INFO ) sum += extended.pages_swapped_out and got sum = 257.4 MiB. But Activity Monitor says: Swap used: 1 GiB and /private/var/vm has 2 GiB swapfiles. Why are these numbers so widely different? Kind regards, Gerriet. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: vmmap
by Gerriet M. Denkmann 26 Mar '12

26 Mar '12

On 24 Mar 2012, at 10:37, comex wrote: > On Fri, Mar 23, 2012 at 5:27 AM, Gerriet M. Denkmann > <gerriet(a)mdenkmann.de> wrote: >> Is the source of the Unix tool vmmap available somewhere? If Yes: where? >> >> If No: given a pid and an address (pointing to the start of a region as returned by vm_region_64) what to do to get answers to the following questions: >> >> - is this region mapped? If Yes: to what file? > > You may want to look at proc_regionfilename from libproc.h (and maybe > its implementation). Thanks. proc_regionfilename() does help a lot. And I would never have found it without your help. Very much appreciated. Kind regards, Gerriet. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Kauth event for write operation
by Quinn "The Eskimo!" 25 Mar '12

25 Mar '12

On 24 Mar 2012, at 07:43, simran charley wrote: > I want to catch each write operation. There's really no supported way to do this in Mac OS X. KAUTH_VNODE_WRITE_DATA authorisation is done when the file is opened and KAUTH_FILEOP_CLOSE notification is done when the file is closed but, for performance reasons, there's no supported way to be authorise or be notified of every single write operation. Share and Enjoy -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Relations, Developer Technical Support, Core OS/Hardware _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Kauth event for write operation
by simran charley 24 Mar '12

24 Mar '12

_______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: vmmap
by comex 24 Mar '12

24 Mar '12

On Fri, Mar 23, 2012 at 5:27 AM, Gerriet M. Denkmann <gerriet(a)mdenkmann.de> wrote: > Is the source of the Unix tool vmmap available somewhere? If Yes: where? > > If No: given a pid and an address (pointing to the start of a region as returned by vm_region_64) what to do to get answers to the following questions: > > - is this region mapped? If Yes: to what file? You may want to look at proc_regionfilename from libproc.h (and maybe its implementation). > - is this region shared with other pid's ? If Yes: which ones? The former question can be sort of answered by vm_region. I don't know any easy way to answer the latter. _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: vmmap
by Stan Sieler 23 Mar '12

23 Mar '12

Re: madvise (thanks to James and Jonas!) ... I'll check into that! >From the name, I suspected that it might not quite meet some needs I've had in the past (and the man page confirms it). I've worked on high-end sorting code that does things like: prefetch (a virtual address, &request_id, NOWAIT) (similar to madvise (..., MADV_WILLNEED), but stronger) (If I wanted to wait for the data to be brought in, I'd pass WAIT instead of NOWAIT ... a prefetch with WAIT is presumably a bit faster/cheaper than getting a page fault on the data.) ... code to do other stuff with other data ... // see if the prefetch is done yet... if (is_prefetch_request_done (request_id, NOWAIT)) ... do something with that data else ... do stuff with other data already in memory I.e., I don't just want to say "hey, I plan to use it" ... I want to say "absolutely start reading it in now, but don't block my code while you do it". Still, a good reference, thanks! Stan _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: vmmap
by James C 23 Mar '12

23 Mar '12

Hi Stan, > (For that matter, let me say "I'm planning to use it in the next few milliseconds... > please fetch it into memory behind my back now, while I do other things"). This one part of your wishes may be possible, but I'm not sure for what proportion of memory, with madvise(..., MADV_WILLNEED). Regards, James. On Sat, Mar 24, 2012 at 8:00 AM, <darwin-kernel-request(a)lists.apple.com> wrote: <snip> > ------------------------------ > > Message: 9 > Date: Fri, 23 Mar 2012 10:44:42 -0700 > From: Stan Sieler <sieler(a)me.com> > To: "Gerriet M. Denkmann" <gerriet(a)mdenkmann.de> > Cc: "darwin-kernel(a)lists.apple.com Kernel" > <darwin-kernel(a)lists.apple.com> > Subject: Re: vmmap > Message-ID: <9772AEC8-AC3B-4746-B3A2-0EFC1943B3CD(a)me.com> > Content-Type: text/plain; CHARSET=US-ASCII > > Re: >>>>> Is the source of the Unix tool vmmap available somewhere? >>> Also, are you trying to get this information for debugging purposes, or to make some meaningful runtime decision on an app that you ship to real users? >> First I want this information for my own entertainment and to educate myself. >> Publishing the resulting app for real users might be something to be thought about in the far future. > > I can sympathize with Gerriet ... I asked similar questions about HP's MPE/iX operating > system, and that's what led me to be able to write tools that were eventually > products used by users *and* HP. > > (With the HP 3000, we had a strong users group ... which helped provide much more > visibility into the kernel and much more interaction with the kernel lab.) > > I find memory (real and virtual) to be one of the hardest things to get good > information about on operating systems other than the now-discontinued MPE/iX. > I like asking questions: is this virtual address valid? Is it readable by me? Writable? > Is this virtual page in memory? Is it dirty? Was it referenced recently? > Is it frozen in memory? (Let me freeze, and later unfreeze it.) > (For that matter, let me say "I'm planning to use it in the next few milliseconds... > please fetch it into memory behind my back now, while I do other things"). > > These are all reasonable questions to ask ... one can easily see how they can lead to > better/faster code in some cases. > > Stan > _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0

Re: vmmap
by Jonas Maebe 23 Mar '12

23 Mar '12

On 23 Mar 2012, at 18:44, Stan Sieler wrote: > (For that matter, let me say "I'm planning to use it in the next few milliseconds... > please fetch it into memory behind my back now, while I do other things"). There is such an API already today: man madvise (I don't know how much is hooked up today though, at least in Mac OS X 10.5 most of those parameters were still no-ops: http://lists.apple.com/archives/PerfOptimization-dev/2009/Apr/msg00024.html ) Jonas _______________________________________________ Do not post admin requests to the list. They will be ignored. Darwin-kernel mailing list (Darwin-kernel(a)lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists… This email sent to site_archiver(a)lists.apple.com

1 0