April 2004 - Darwin-kernel - lists.apple.com

Re: Power Manager Question...
by Jerry LeVan 30 Apr '04

30 Apr '04

Well the power manager needs a bit of help, if the mac is awakened by the "wake" event it will go back to sleep in just a couple of minutes... See http://homepage.mac.com/levanj/Cocoa for my rube goldberg schema for getting cron to fire on time with enough time :) Jerry On Apr 30, 2004, at 11:18 AM, Joey Echeverria wrote: I have a program that uses the IO Power Manager API to queue a wake request to fire about a minute and 15 seconds before cron is scheduled to fire Could this become the default behavior? I know I've missed many a cron job due to a sleeping laptop. Thanks, -Joey Echeverria _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: Power Manager Question...
by Ethan Bold 28 Apr '04

28 Apr '04

Hey Jerry- Good question. I suspect this is due to a PM timer issue, where we don't fully reset the sleep timer unless the system was awoken by a key press (or clamshell open or power button press). You probably need to fake a keypress, as you suspected. The AppleScript fake click MIGHT work - I don't know any AppleScript though, and I don't know how their mouseClicks filter through the system. If you can send me a script I can check whether it tickles the PM timers or not. Alternatively, this C program will fake a key press: // cc -o keypress keypress.c -framework Carbon #include <Carbon/Carbon.h> main() { UpdateSystemActivity(OverallAct); return 0; } -ethan On Apr 28, 2004, at 12:46 PM, Jerry LeVan wrote: > This might not be the best forum for this question but I don't > know quite where to look for an solution... > > I have a program that uses the IO Power Manager API to queue a > wake request to fire about a minute and 15 seconds before cron is > scheduled > to fire ( at 3:15 in the morning). > > Examining the system log I see that the system goes back to sleep > in about 3~4 minutes after the wakeup call. The Energy Saver panel > is set to sleep in 15 minutes of inactivity. > > Why is there such a big discrepancy in the time? It appears to me > that a big backup might get timed out by the sleep mechanism... > > I suspect there might be some watchdog timer set every time a key is > pressed or the mouse moves, at 3:15 there is no one about to jiggle > the keyboard/mouse. > > Is there a software solution to this problem...would an Applescript > system event fake click do the job? > > I guess one solution would be to point a photosensor at the screen so > that when the wakeup occurs, the sensor could cut power to a > electromagnet > holding a ball above the key board but that seems like overkill :) > > Jerry > _______________________________________________ > darwin-kernel mailing list | darwin-kernel(a)lists.apple.com > Help/Unsubscribe/Archives: > http://www.lists.apple.com/mailman/listinfo/darwin-kernel > Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Power Manager Question...
by Jerry LeVan 28 Apr '04

28 Apr '04

This might not be the best forum for this question but I don't know quite where to look for an solution... I have a program that uses the IO Power Manager API to queue a wake request to fire about a minute and 15 seconds before cron is scheduled to fire ( at 3:15 in the morning). Examining the system log I see that the system goes back to sleep in about 3~4 minutes after the wakeup call. The Energy Saver panel is set to sleep in 15 minutes of inactivity. Why is there such a big discrepancy in the time? It appears to me that a big backup might get timed out by the sleep mechanism... I suspect there might be some watchdog timer set every time a key is pressed or the mouse moves, at 3:15 there is no one about to jiggle the keyboard/mouse. Is there a software solution to this problem...would an Applescript system event fake click do the job? I guess one solution would be to point a photosensor at the screen so that when the wakeup occurs, the sensor could cut power to a electromagnet holding a ball above the key board but that seems like overkill :) Jerry _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: SysV Shared Memory Problems
by Simon Patience 26 Apr '04

26 Apr '04

Yes, if you do not call shmctl() with IPC_RMID, then the id will continue to remain valid after you call shmdt() on that id. This is how processes rendezvous with the shared memory segment over time. The life time of the id is either until you remove it with IPC_RMID or until you reboot. Simon. On Apr 25, 2004, at 2:21 AM, Adrian Bool wrote: Hi Alec, I've had a play with shmget out of interest - boring Sunday! ;-) I do not have any problem with allocating a shared memory area larger than 1024B. The only time I got an error back from shmget was as so, * I used my test program to create a 4*1024 block to both write to that area and - with a different process - read. No probs. * I then wanted to see how big I could make it, so changed the size to 512 *1024. * At this point I received an error, 'shmget failed: Invalid argument'. * What I did *NOT* do was change the key - it was just '1'. If I then changed the key to something obvious like '2' I could allocate the 512*1024 memory block successfully. Could this possibly be your problem...? You have initially created a 1024 byte block under key 'X'. Calling your program again with the same size block size and a key of 'X' gives no trouble as the system merely returns the same block. Attempting to create any other sized block with the same key is invalid and so the system does not return the block you had previously created but instead gives you your error? So a change of key may help you? I suppose I should be destroying my shared block and may not then see the above. Shall see how to do that... regards, aid On 25 Apr 2004, at 05:53, Alec Ellsworth wrote: Hi All, I'm trying to port one of my apps from FreeBSD to OSX and am having a problem during runtime. This app is a processor emulator for a proprietary architecture and uses SysV shared memory for IPC. On FreeBSD 4 and 5 (with SysV shared memory turned on) I don't have a problem allocating large amounts of shared space, but under OSX I'm getting a -1 returned if I try to allocate more than 1024 _bytes_! Here's the code line that returns false: id=shmget(key,kreq*1024, 0777 | IPC_CREAT); Any time kreq>1, it fails. I've made sure I'm set up for larger spaces via sysctl, but I must be missing something. My app only allocates the block once and there is no other shm activity before this point. >From sysctl -a: kern.sysv.shmmax: 167772160 kern.sysv.shmmin: 1 kern.sysv.shmmni: 32 kern.sysv.shmseg: 8 kern.sysv.shmall: 65536 Thanks for any input. So far I absolutely love working in Xcode! -Alec Ellsworth Alec Ellsworth Resilient Networks, LLC. 3701 W. Burnham Street Suite A Milwaukee, WI 53215 alec(a)resilientnets.net _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: SysV Shared Memory Problems
by Adrian Bool 25 Apr '04

25 Apr '04

Hi Alec, I've had a play with shmget out of interest - boring Sunday! ;-) I do not have any problem with allocating a shared memory area larger than 1024B. The only time I got an error back from shmget was as so, * I used my test program to create a 4*1024 block to both write to that area and - with a different process - read. No probs. * I then wanted to see how big I could make it, so changed the size to 512 *1024. * At this point I received an error, 'shmget failed: Invalid argument'. * What I did *NOT* do was change the key - it was just '1'. If I then changed the key to something obvious like '2' I could allocate the 512*1024 memory block successfully. Could this possibly be your problem...? You have initially created a 1024 byte block under key 'X'. Calling your program again with the same size block size and a key of 'X' gives no trouble as the system merely returns the same block. Attempting to create any other sized block with the same key is invalid and so the system does not return the block you had previously created but instead gives you your error? So a change of key may help you? I suppose I should be destroying my shared block and may not then see the above. Shall see how to do that... regards, aid On 25 Apr 2004, at 05:53, Alec Ellsworth wrote: Hi All, I'm trying to port one of my apps from FreeBSD to OSX and am having a problem during runtime. This app is a processor emulator for a proprietary architecture and uses SysV shared memory for IPC. On FreeBSD 4 and 5 (with SysV shared memory turned on) I don't have a problem allocating large amounts of shared space, but under OSX I'm getting a -1 returned if I try to allocate more than 1024 _bytes_! Here's the code line that returns false: id=shmget(key,kreq*1024, 0777 | IPC_CREAT); Any time kreq>1, it fails. I've made sure I'm set up for larger spaces via sysctl, but I must be missing something. My app only allocates the block once and there is no other shm activity before this point. >From sysctl -a: kern.sysv.shmmax: 167772160 kern.sysv.shmmin: 1 kern.sysv.shmmni: 32 kern.sysv.shmseg: 8 kern.sysv.shmall: 65536 Thanks for any input. So far I absolutely love working in Xcode! -Alec Ellsworth Alec Ellsworth Resilient Networks, LLC. 3701 W. Burnham Street Suite A Milwaukee, WI 53215 alec(a)resilientnets.net _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

SysV Shared Memory Problems
by Alec Ellsworth 25 Apr '04

25 Apr '04

Hi All, I'm trying to port one of my apps from FreeBSD to OSX and am having a problem during runtime. This app is a processor emulator for a proprietary architecture and uses SysV shared memory for IPC. On FreeBSD 4 and 5 (with SysV shared memory turned on) I don't have a problem allocating large amounts of shared space, but under OSX I'm getting a -1 returned if I try to allocate more than 1024 _bytes_! Here's the code line that returns false: id=shmget(key,kreq*1024, 0777 | IPC_CREAT); Any time kreq>1, it fails. I've made sure I'm set up for larger spaces via sysctl, but I must be missing something. My app only allocates the block once and there is no other shm activity before this point. >From sysctl -a: kern.sysv.shmmax: 167772160 kern.sysv.shmmin: 1 kern.sysv.shmmni: 32 kern.sysv.shmseg: 8 kern.sysv.shmall: 65536 Thanks for any input. So far I absolutely love working in Xcode! -Alec Ellsworth Alec Ellsworth Resilient Networks, LLC. 3701 W. Burnham Street Suite A Milwaukee, WI 53215 alec(a)resilientnets.net _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: raising the stacksize limit
by Steven Bytnar 23 Apr '04

23 Apr '04

Here's my guess... On PowerPC architectures, virtual memory is broken into segments. Each segment is 256MB large. The code that lays out each processes memory address space probably is probably coded to create one and only one stack segment. You'll probably need to modify something around that mechanism. I'm not sure, but I imagine there could be some weird virtual memory issues regarding cross-segment stack boundaries which would be why this isn't implemented. *shrug* Just another guess. --Steve On Fri, Apr 23, 2004 at 11:04:19AM -0600, Nicolas Bock wrote: > Hello list, > > I was trying to raise the stacksize limit of the OS X 10.3.3 kernel by > changing > > #define MAXSSIZ (64*1024*1024) > > to > > #define MAXSSIZ (RLIM_INFINITY) > > in the file > > xnu-517.3.15/bsd/ppc/vmparam.h > > The new kernel reports the hard limit for stacksize to be unlimited as > I had hoped, but I am still not able to raise the actual stacksize > limit to anything beyond 256 MB. Does anyone know what the reason for > that could be? Do I have to change something else? > > Thanks, nick > _______________________________________________ > darwin-kernel mailing list | darwin-kernel(a)lists.apple.com > Help/Unsubscribe/Archives: > http://www.lists.apple.com/mailman/listinfo/darwin-kernel > Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

raising the stacksize limit
by Nicolas Bock 23 Apr '04

23 Apr '04

Hello list, I was trying to raise the stacksize limit of the OS X 10.3.3 kernel by changing #define MAXSSIZ (64*1024*1024) to #define MAXSSIZ (RLIM_INFINITY) in the file xnu-517.3.15/bsd/ppc/vmparam.h The new kernel reports the hard limit for stacksize to be unlimited as I had hoped, but I am still not able to raise the actual stacksize limit to anything beyond 256 MB. Does anyone know what the reason for that could be? Do I have to change something else? Thanks, nick _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

Re: compiling the kernel problems
by Nicolas Bock 23 Apr '04

23 Apr '04

Thanks Felix, that helped. I was able to compile the kernel and rebooted successfully. Thanks again, nick On Apr 23, 2004, at 8:09 AM, Felix Kronlage wrote: On Thu, Apr 22, 2004 at 06:02:37PM -0700, Justin Walker wrote: Shantonu has instructions on opendarwin.org's site. Search the darwin-development list archives for mail with the subject 'Building xnu', from June of last year. There may also be pointers at that site (<http://www.opendarwin.org>). the most recent notes around is the 'building xnu' howto over at OpenDarwin.org. While it was written around 10.2.x (= darwin 6.6.x) it still applies for 10.3.x (darwin 7.x). The difference is, that now also the kext_tools project belongs to the toolchain (which the orgininal poster stated that he was missing). An easy way to get 'kextsymboltool' is to grab it off the darwin 7.0.1 install (it should be in /usr/local/bin). -fkr -- gpg-fingerprint: 076E 1E87 3E05 1C7F B1A0 8A48 0D31 9BD3 D9AC 74D0 Felix Kronlage | FKR-RIPE | http://www.hazardous.org/fkr/ fkr(a){bytemine.net|opendarwin.org} | http://opendarwin.org/~fkr/ [demime 0.98b removed an attachment of type application/pgp-signature] _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored. _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0

darwin sources/xnu
by Matt Jaffa 23 Apr '04

23 Apr '04

Hi, I am a little confused, I have been looking in at kern_sysctl.c, and kern_newsysctl.c in the kern folder of the xnu darwin source to understand how kernel_sysctl(......) works. and in both files it is defined there, so which kernel_sysctl is the one that my KEXT project is going to use, I am assuming kernel_sysctl in the kern_newsysctl.c since it has new in it. What if I want to use the old one, cause it seems that the kernel_sysctl in the old one would work, right now kernel_sysctl is always returning error 2, ENOENT. Matt _______________________________________________ darwin-kernel mailing list | darwin-kernel(a)lists.apple.com Help/Unsubscribe/Archives: http://www.lists.apple.com/mailman/listinfo/darwin-kernel Do not post admin requests to the list. They will be ignored.

1 0