November 2015 - Darwin-kernel

Re: OS X 10.11 sources
by Brian Bechtel 23 Nov '15

23 Nov '15

Unfortunately, developer support is the only way you are likely to get an answer. https://developer.apple.com/contact/ This week, Apple Cupertino is shut down for a holiday break. So any emails you send will not be seen until Monday. > On Nov 23, 2015, at 6:53 AM, Nathan Hjelm <hjelmn(a)mac.com> wrote: > > I was wondering the same thing. I am trying to track down a bug in IOUSBFamily but the source for 10.11 still hasn’t been posted. > > -Nathan > >> On Nov 23, 2015, at 7:44 AM, Andreas Fink <afink(a)list.fink.org> wrote: >> >> 10.11 has been released over 4 months ago. >> Whats the reason the sources have still not been released?? >> >> _______________________________________________ >> 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/hjelmn%40mac.com >> >> This email sent to hjelmn(a)mac.com > > > _______________________________________________ > 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/brian.bechtel%40gmail… > > This email sent to brian.bechtel(a)gmail.com _______________________________________________ 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: OS X 10.11 sources
by Nathan Hjelm 23 Nov '15

23 Nov '15

I was wondering the same thing. I am trying to track down a bug in IOUSBFamily but the source for 10.11 still hasn’t been posted. -Nathan > On Nov 23, 2015, at 7:44 AM, Andreas Fink <afink(a)list.fink.org> wrote: > > 10.11 has been released over 4 months ago. > Whats the reason the sources have still not been released?? > > _______________________________________________ > 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/hjelmn%40mac.com > > This email sent to hjelmn(a)mac.com _______________________________________________ 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: superpages?
by Stan Sieler 05 Nov '15

05 Nov '15

Re: > Super pages requires that the backing 2MB be physically contiguous. After a system has been up for a while, wired pages may fragment physical memory and prevent 2MB of physically contiguous pages from being found. Try running the test program after a reboot to verify if this is the case. Hmm…would be interesting to have a utility that would display a map of physical memory, indicating the wired pages and who owns them. I’d have fun with that! Sounds like OS X needs the same enhancement request I made for MPE/iX some years back, although hopefully with a different result (MPE/iX (and the HP 3000) were killed after I made my request, but I’m sure cause and effect :)) Enhancement: When code requests wired / permanent memory, the OS will strive to make wired memory contiguous with existing wired memory, to avoid fragmenting physical memory. One possible approach could be: 1. find the address just after the last satisfied wired request; 2. find the address of the first non-wired address area that would work. if #2 is lower than #1, #2 is used. (this could happen if a prior wired chunk was released) (If #2 + length extends past prior #1, remember it as new #1, of course) else sufficient memory is kicked out starting at #1, then #1 is used, and the end address is recorded for future use in #1. (Apologies for any OS X terminology misuse … I work on too many OSes!) (I had needed to get 3 pages of contiguous physical space (wired) allocated on MPE/iX (probably the only person outside of of HP to want to do this), and the space had to be in the first 4 GB of physical space … turns out the OS couldn’t reliably do it after the system had been up for awhile :) thanks, Stan Sieler _______________________________________________ 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: superpages?
by Nathan Hjelm 05 Nov '15

05 Nov '15

> On Nov 4, 2015, at 11:25 PM, Shantonu Sen <ssen(a)apple.com> wrote: > > The functionality is not for general purpose use. Ok, thats unfortunate. Was hoping for something at least as usable as huge pages is in Linux. Is there any chance superpages could be made more useful in a future xnu version? > Super pages requires that the backing 2MB be physically contiguous. After a system has been up for a while, wired pages may fragment physical memory and prevent 2MB of physically contiguous pages from being found. Try running the test program after a reboot to verify if this is the case. Yup, runs fine in the first 5 or so minutes after a reboot then starts failing again. This is on a machine with 16 GB of memory so I didn’t expect the memory used for superpages to be exhausted that fast. -Nathan Hjelm _______________________________________________ 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: superpages?
by Shantonu Sen 05 Nov '15

05 Nov '15

The functionality is not for general purpose use. Super pages requires that the backing 2MB be physically contiguous. After a system has been up for a while, wired pages may fragment physical memory and prevent 2MB of physically contiguous pages from being found. Try running the test program after a reboot to verify if this is the case. Shantonu > On Nov 4, 2015, at 6:17 PM, Nathan Hjelm <hjelmn(a)mac.com> wrote: > > Are superpages still supported in OS X. I tried to run the test in xnu (http://www.opensource.apple.com/source/xnu/xnu-2782.40.9/tools/tests/superp…) but it fails pretty much every test: > > ➜ /tmp ./testsp > Running 14 tests: > Test #1 "allocate one page anywhere"...FAILED! (mach_vm_allocate() returned 3) > Test #2 "deallocate a page"...FAILED! (skipped deallocation) > Test #3 "allocate a SIZE_ANY page anywhere"...FAILED! (mach_vm_allocate() returned 3) > Test #4 "allocate one page at a fixed address"...FAILED! (mach_vm_allocate() returned 3) > Test #5 "allocate one page at an unaligned fixed address"...OK > Test #6 "deallocate sub-page"...FAILED! (mach_vm_allocate() returned 3) > Test #7 "allocate already allocated subpage"...FAILED! (mach_vm_allocate() returned 3) > Test #8 "wire a page"...FAILED! (mach_vm_allocate() returned 3) > Test #9 "unwire a page"...FAILED! (mach_vm_allocate() returned 3) > Test #10 "make page readonly"...FAILED! (mach_vm_allocate() returned 3) > Test #11 "make sub-page readonly"...FAILED! (mach_vm_allocate() returned 3) > Test #12 "file I/O"...FAILED! (mach_vm_allocate (1)() returned 3) > Test #13 "mmap()"...FAILED! (mmap()) > Test #14 "fork"...FAILED! (mach_vm_allocate() returned 3) > > Ran this on 10.11.1. Haven’t had a chance to try other OS X versions. > > -Nathan Hjelm > _______________________________________________ > 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/ssen%40apple.com > > This email sent to ssen(a)apple.com _______________________________________________ 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

superpages?
by Nathan Hjelm 05 Nov '15

05 Nov '15

Are superpages still supported in OS X. I tried to run the test in xnu (http://www.opensource.apple.com/source/xnu/xnu-2782.40.9/tools/tests/superp…) but it fails pretty much every test: ➜ /tmp ./testsp Running 14 tests: Test #1 "allocate one page anywhere"...FAILED! (mach_vm_allocate() returned 3) Test #2 "deallocate a page"...FAILED! (skipped deallocation) Test #3 "allocate a SIZE_ANY page anywhere"...FAILED! (mach_vm_allocate() returned 3) Test #4 "allocate one page at a fixed address"...FAILED! (mach_vm_allocate() returned 3) Test #5 "allocate one page at an unaligned fixed address"...OK Test #6 "deallocate sub-page"...FAILED! (mach_vm_allocate() returned 3) Test #7 "allocate already allocated subpage"...FAILED! (mach_vm_allocate() returned 3) Test #8 "wire a page"...FAILED! (mach_vm_allocate() returned 3) Test #9 "unwire a page"...FAILED! (mach_vm_allocate() returned 3) Test #10 "make page readonly"...FAILED! (mach_vm_allocate() returned 3) Test #11 "make sub-page readonly"...FAILED! (mach_vm_allocate() returned 3) Test #12 "file I/O"...FAILED! (mach_vm_allocate (1)() returned 3) Test #13 "mmap()"...FAILED! (mmap()) Test #14 "fork"...FAILED! (mach_vm_allocate() returned 3) Ran this on 10.11.1. Haven’t had a chance to try other OS X versions. -Nathan Hjelm _______________________________________________ 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