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

Re: Max udp buffer size
by Andrew Gallatin 30 Apr '12

30 Apr '12

On 04/30/12 09:46, Matthew Wallace wrote: Worked like a champ - thanks so much for your help! As a curiousity, I still haven't been able to figure out where the value of 5591040 came from. If the value was really 1/16 of 64MB that would give 4MB as the max recv buffers - any ideas how they arrived at the seemingly random value of 5591040? I was wondering the same thing. I'd assumed you had changed something on your end away from the defaults first.. Drew _______________________________________________ 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: Max udp buffer size
by Matthew Wallace 30 Apr '12

30 Apr '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: Max udp buffer size
by Andrew Gallatin 30 Apr '12

30 Apr '12

On 04/29/12 20:53, Matthew Wallace wrote: Hoping this is the right forum for this question - I've noticed a peculiar limit when setting net.inet.udp.recvspace using sysctl. This value cannot be any larger than 5591040 bytes - any value larger than that gives the error message "Result too large". This is on a MacPro running OS X 10.7.3. Can anybody give me any clue as to where this limitation comes from and if there's any way around it? A while back (10.6?) Apple made socket buffer sizes conditional on the amount of number of mbuf clusters. The value you're trying to adjust is limited by sb_max. That, in turn, is set by kern.ipc.maxsockbuf. That is limited to 1/16 of the amount of memory allocated to network buffers. I think the goal was to prevent one or two connections from being able to starve the entire machine. The easiest fix is to increase the number of mbuf clusters using the ncl=??? argument. The default mbuf cluster pool size is 64MB (32768 2KB clusters), and the maxsockbuf limit is indeed 1/16 of that: root@macpro03:~# sysctl -a | egrep nmbcl\|maxsockb kern.ipc.maxsockbuf: 4194304 kern.ipc.nmbclusters: 32768 To fix this, tell the system to use more than the default 32768 mbufclusters via the ncl boot arg: root@macpro02:~# nvram boot-args="ncl=131072" root@macpro02:~# shutdown -r now When the machine comes back up, it has more mbuf clusters, and will allow you to raise kern.ipc.maxsockbuf: root@macpro02:~# sysctl -w kern.ipc.maxsockbuf=16777216 kern.ipc.maxsockbuf: 8388608 -> 16777216 root@macpro02:~# sysctl -a | egrep nmbcl\|maxsockb kern.ipc.maxsockbuf: 16777216 kern.ipc.nmbclusters: 131072 Now you can finally increase udp recvspace: root@macpro02:~# sysctl -w net.inet.udp.recvspace=8388608 net.inet.udp.recvspace: 42080 -> 8388608 Drew _______________________________________________ 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

Max udp buffer size
by Matthew Wallace 30 Apr '12

30 Apr '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: libtop.c: two questions
by Jimgreen8＠me.com 29 Apr '12

29 Apr '12

Jimgreen On Apr 28, 2012, at 19:55, "Gerriet M. Denkmann" <gerriet(a)mdenkmann.de> wrote: > Reading <http://www.opensource.apple.com/source/top/top-73/libtop.c> I got two questions: > > > 1. > static void > libtop_p_fw_scan > (task_t task, mach_vm_address_t region_base, mach_vm_size_t region_size) { > [...] > kr = mach_vm_region_recurse(task, &addr, &size, &depth,(vm_region_recurse_info_t)&sinfo, &count); > [...] > switch (sinfo.share_mode) { > case SM_SHARED: > case SM_COW: > case SM_TRUESHARED: > if (sinfo.max_protection & VM_PROT_EXECUTE) { > // CODE > code += sinfo.pages_resident; > tsamp.fw_count++; > } else if (sinfo.max_protection & VM_PROT_WRITE) { > // DATA > data += sinfo.pages_resident; > } else { > // LINKEDIT > linkedit += sinfo.pages_resident; > } > break; > } > [...] > > But when I look at the results of mach_vm_region_recurse I find that protection/max_protection of: > __TEXT is r-x/r-x > __DATA, __OBJC have rw-/rwx > __LINKEDIT has r--/rwx > > So the above code (which checks max_protection) either does make no sense, or I am misunderstanding something. > > > 2. > In libtop_update_vm_regions (task_t task, libtop_pinfo_t* pinfo) > I see: > mach_vm_region(...) > case SM_PRIVATE: > rprvt += info.private_pages_resident * pagesize; > rprvt += info.shared_pages_resident * pagesize; > > But looking at the results of mach_vm_region(VM_REGION_TOP_INFO) I see that while private_pages_resident, shared_pages_resident ≤ size, quite often private_pages_resident + shared_pages_resident > size. > > Does this code really count rprvt correctly? > > > 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/jimgreen8%40me.com > > This email sent to jimgreen8(a)me.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

libtop.c: two questions
by Gerriet M. Denkmann 28 Apr '12

28 Apr '12

Reading <http://www.opensource.apple.com/source/top/top-73/libtop.c> I got two questions: 1. static void libtop_p_fw_scan (task_t task, mach_vm_address_t region_base, mach_vm_size_t region_size) { [...] kr = mach_vm_region_recurse(task, &addr, &size, &depth,(vm_region_recurse_info_t)&sinfo, &count); [...] switch (sinfo.share_mode) { case SM_SHARED: case SM_COW: case SM_TRUESHARED: if (sinfo.max_protection & VM_PROT_EXECUTE) { // CODE code += sinfo.pages_resident; tsamp.fw_count++; } else if (sinfo.max_protection & VM_PROT_WRITE) { // DATA data += sinfo.pages_resident; } else { // LINKEDIT linkedit += sinfo.pages_resident; } break; } [...] But when I look at the results of mach_vm_region_recurse I find that protection/max_protection of: __TEXT is r-x/r-x __DATA, __OBJC have rw-/rwx __LINKEDIT has r--/rwx So the above code (which checks max_protection) either does make no sense, or I am misunderstanding something. 2. In libtop_update_vm_regions (task_t task, libtop_pinfo_t* pinfo) I see: mach_vm_region(...) case SM_PRIVATE: rprvt += info.private_pages_resident * pagesize; rprvt += info.shared_pages_resident * pagesize; But looking at the results of mach_vm_region(VM_REGION_TOP_INFO) I see that while private_pages_resident, shared_pages_resident ≤ size, quite often private_pages_resident + shared_pages_resident > size. Does this code really count rprvt correctly? 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: setting kernel shared memory on 10.7.3
by Andrew Barinov 28 Apr '12

28 Apr '12

Andrew Andrew _______________________________________________ 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: setting kernel shared memory on 10.7.3
by George Plymale 28 Apr '12

28 Apr '12

Andrew _______________________________________________ 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

setting kernel shared memory on 10.7.3
by Andrew Barinov 28 Apr '12

28 Apr '12

Andrew _______________________________________________ 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: How to get nbr of resident pages
by Gerriet M. Denkmann 26 Apr '12

26 Apr '12

On 25 Apr 2012, at 05:42, comex wrote: > On Sat, Apr 21, 2012 at 3:10 AM, Gerriet M. Denkmann > <gerriet(a)mdenkmann.de> wrote: >> I thought that vmmeter.v_wire_count + vmmeter .v_active_count would be the answer. >> >> So I tried: >> int mib[] = { CTL_VM, VM_METER }; >> [...] >> int res = sysctl(mib, miblen, &vmmeter, &len, NULL, 0); >> >> But got: errno = 2 = "No such file or directory" >> >> So: How do I get the number of resident pages of all processes? > > VM_METER doesn't seem to be implemented in xnu, but top, which reports > this information, uses a different call: > > kr = host_statistics(libtop_port, HOST_VM_INFO, > (host_info_t)&tsamp->vm_stat, &count); > > - http://www.opensource.apple.com/source/top/top-73/libtop.c Thanks! This is exactly what I needed. 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