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

Kernel Panics in my NKE
by Matt Jaffa 22 Jun '04

22 Jun '04

Hi, I was wondering if there is a limit within a NKE of how much data can be allocated. For instance my NKE is working great until a packet with a length of 65000+ come in and when this line is reached unsigned char buffer[sizeofpacket] it crashes cause it is too big I guess, usually packets are only 2 - 1000 so this has no problem, when I remove that line of code my NKE never crashes. So what is the limit of what I can do? I mean the size of the unsigned char buffer[size]. 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

Re: File Monitoring under Mac OS X
by Jonas Maebe 21 Jun '04

21 Jun '04

On 21 jun 2004, at 22:03, Krishna Monian wrote: What is "the complete path"? In world with hard links, one file can have countless "complete paths". What I mean is the actual path for a given file and not just a relative path. /Users/admin/.... and so on And what I mean is that a file may have more than one actual path and location, and that it wouldn't surprise me if inside the kernel there is no notion of "the path" to a file, only some location on disk (possibly together with whatever the user specified, but as you noted that does not have to be a complete path) Jonas _______________________________________________ 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: File Monitoring under Mac OS X
by Krishna Monian 21 Jun '04

21 Jun '04

What I mean is the actual path for a given file and not just a relative path. /Users/admin/.... and so on Thanks --- Jonas Maebe <jonas(a)zeus.ugent.be> wrote: > On 21 jun 2004, at 19:28, Krishna Monian wrote: > > > Would you happen to know whether there is any way > in > > which I can get the complete path from the system. > > What is "the complete path"? In world with hard > links, one file can > have countless "complete paths". > > > Jonas > _______________________________________________ > 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. > > __________________________________ Do you Yahoo!? New and Improved Yahoo! Mail - Send 10MB messages! http://promotions.yahoo.com/new_mail _______________________________________________ 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: Kernel Panic when writing to disk [update]
by Justin Walker 21 Jun '04

21 Jun '04

On Jun 21, 2004, at 10:05, Matthieu Kopp wrote: Hi, I admit it was a very bad idea, .... the rational behind it was to develop a standalone widget that would give an up-to-date indication of the amount of free disk space without having to know about the processes performing disk operations. This is why I used a timer (OK, the polling frequency was very high). Gak! I can appreciate the desire, I guess, but it seems much more efficient to have the widget wait for a request, and *then* get up-to-date information. With polling, your information is only as good as the last update. You have to have a high frequency to assure that the information you return is (more or less) accurate, and then you end up paying that high price for the information. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | Some people have a mental | horizon of radius zero, and | call it their point of view. | -- David Hilbert *--------------------------------------*-------------------------------* _______________________________________________ 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: File Monitoring under Mac OS X
by Jonas Maebe 21 Jun '04

21 Jun '04

On 21 jun 2004, at 19:28, Krishna Monian wrote: Would you happen to know whether there is any way in which I can get the complete path from the system. What is "the complete path"? In world with hard links, one file can have countless "complete paths". Jonas _______________________________________________ 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: File Monitoring under Mac OS X
by Krishna Monian 21 Jun '04

21 Jun '04

Would you happen to know whether there is any way in which I can get the complete path from the system. Thanks --- Marek Kozubal <marek(a)portents.com> wrote: > On Fri, 18 Jun 2004, Krishna Monian wrote: > > > One question I have on fs_usage is that, if you > access > > a file from a bash shell (say more <filename>) > then it > > doesn't report the entire path, but only the > filename. > > Would you happen to know the reason behind this > and > > whether it can be changed? > > I would imagine that fs_usage is probably printing > the string that exists > in cnp->cn_pathbuf. When used from a Carbon > application (and probably > Cocoa too) you'll see /.vol/<10 digit number/<CNID>. > While from bash and > other posix shells, the current directory is stored > as a vnode, and so all > you get is what the user typed into the command > line. So if he typed > "more foo", all you'd get is foo. If they typed > "/Volumes/drive/foobar" > you'd see /Volumes/drive/foobar". > > Inside the Kernel, you generally don't reference > things by full path > names, its *a* parent vnode (there can be multiple > different parents to a > file) and a name, which is converted to a vnode. > > I will note again, that this is a complete guess, > but if you're in kernel > land watching file system operations this is the > kind of information you'd > have to work with. > > --- > Marek Kozubal > marek(a)portents.com > __________________________________ Do you Yahoo!? New and Improved Yahoo! Mail - Send 10MB messages! http://promotions.yahoo.com/new_mail _______________________________________________ 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: Kernel Panic when writing to disk [update]
by Matthieu Kopp 21 Jun '04

21 Jun '04

Hi, I admit it was a very bad idea, .... the rational behind it was to develop a standalone widget that would give an up-to-date indication of the amount of free disk space without having to know about the processes performing disk operations. This is why I used a timer (OK, the polling frequency was very high). Mat On 21 Jun 2004, at 17:03, Justin Walker wrote: On Jun 21, 2004, at 1:44, Matthieu Kopp wrote: Hi, After spending a few more hours on the problem, I finally found what was causing the panic...and more importantly for me, how to avoid it. I am still posting a bug to Apple, because as you mentioned it, this should never happen. Be sure to add this to your bug report :-} What triggers the panic in my application is the following: my application is multi-threaded and as I mentionned, I have one thread saving data at regular intervals (using writeToFile:atomically of NSData). Another thread is triggered by a timer and is checking the amount of free disk space every second using : Maybe I'm unclear on what is going on, but it seems to be a terrific waste of resources to check once per second, on a thread devoted to that task, for freespace. Wouldn't it be better to check before you do the write? In general, polling is A Bad Thing, except in dedicated devices (and then, only when the polling doesn't involve "going external"). There are always exceptions, but you should try to design polling out of your application. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | It's not whether you win or lose... | It's whether *I* win or lose. *-------------------------------------- *-------------------------------* _______________________________________________ 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: Kernel Panic when writing to disk [update]
by Eric Seidel 21 Jun '04

21 Jun '04

Do you have a radar # for this one? -eric On Jun 21, 2004, at 1:44 AM, Matthieu Kopp wrote: Hi, After spending a few more hours on the problem, I finally found what was causing the panic...and more importantly for me, how to avoid it. I am still posting a bug to Apple, because as you mentioned it, this should never happen. What triggers the panic in my application is the following: my application is multi-threaded and as I mentionned, I have one thread saving data at regular intervals (using writeToFile:atomically of NSData). Another thread is triggered by a timer and is checking the amount of free disk space every second using : NSFileManager *fm = [NSFileManager defaultManager]; NSDictionary *fsattrs = [fm fileSystemAttributesAtPath:@"/"]; If the data being written is big enough so that it takes a few seconds to write, I can see that the call to writeToFile:atomically:(NSData) never returns. The thread performing the write operation has become uninterruptible and stays so until the system crash. I assume it is the call to [fm fileSystemAttributesAtPath:@"/"] that interferes with the write operation and causes the write operation to fail. The workaround for me is simply to avoid doing that call to fileSystemAttributesAtPath when the program is in the middle of a save operation. I assume that knowledgeable readers of that list will know why the observed thread behavior causes the system to panic. Thanks for the replies. Matthieu On 18 Jun 2004, at 21:32, Shawn Erickson wrote: On Jun 18, 2004, at 10:46 AM, Matthieu Kopp wrote: Hi, I am posting this question to both groups as it involves both drivers and kernel knowledge. I am working on a piece of COCOA code that writes data to disk at regular intervals. The data is written using a NSData object using the writeToFile:atomically method.The size of the file produced during the write operation varies between 100Kb and 5Mb and I typically produce 5 files/minute. Very often, I am getting a kernel panic during the write operation. It can happen after just 2 minutes of operation or at a later stage, say after an hour. It is much easier to get the panic when the writing is performed by a thread that is not the main thread. (I typically do the writing using a ThreadWorker). My test system is a Powerbook G4 (550Mhz) with 1.9Gb of free disk space. I *cannot* reproduce this problem on my G5 2x2Ghz. A sample panic log is shown below. As a side note regardless of what you do in your application you should not be able to trigger a kernel panic, especially if using standard IO services like you are, no matter how hard you try. File a bug report with the panic information, etc. ... <http://developer.apple.com/bugreporter/index.html> -Shawn _______________________________________________ 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: Kernel Panic when writing to disk [update]
by Justin Walker 21 Jun '04

21 Jun '04

On Jun 21, 2004, at 1:44, Matthieu Kopp wrote: Hi, After spending a few more hours on the problem, I finally found what was causing the panic...and more importantly for me, how to avoid it. I am still posting a bug to Apple, because as you mentioned it, this should never happen. Be sure to add this to your bug report :-} What triggers the panic in my application is the following: my application is multi-threaded and as I mentionned, I have one thread saving data at regular intervals (using writeToFile:atomically of NSData). Another thread is triggered by a timer and is checking the amount of free disk space every second using : Maybe I'm unclear on what is going on, but it seems to be a terrific waste of resources to check once per second, on a thread devoted to that task, for freespace. Wouldn't it be better to check before you do the write? In general, polling is A Bad Thing, except in dedicated devices (and then, only when the polling doesn't involve "going external"). There are always exceptions, but you should try to design polling out of your application. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | It's not whether you win or lose... | It's whether *I* win or lose. *--------------------------------------*-------------------------------* _______________________________________________ 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: Kernel Panic when writing to disk [update]
by Markus Hitter 21 Jun '04

21 Jun '04

Congrats for discovering the bug. Am 21.06.2004 um 10:44 schrieb Matthieu Kopp: Another thread is triggered by a timer and is checking the amount of free disk space every second ... It's never a good idea to access the disk in regular intervals, IMHO[1]. This will avoid stopping the disk when it's otherwise unused. Annoying, especially on i/PowerBooks. >From what you told us about your goals, it's fully sufficient to check the disk's free space at the beginning of each write operation. my $0.02 Markus [1] Printer Utility serves as a poor example here, it checks for incoming print jobs each second. Drives me nuts if this app is running somewhere in the background (e.g. due to a failed print operation) and the iBook won't go to sleep for no visible reason. - - - - - - - - - - - - - - - - - - - Dipl. Ing. Markus Hitter http://www.jump-ing.de/ _______________________________________________ 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