March 2003 - Darwin-kernel - lists.apple.com

Re: realloc
by Eric Grant 18 Mar '03

18 Mar '03

On Mon, 17 Mar 2003 09:11:53 +0100, Jonas Maebe <jonas(a)zeus.rug.ac.be> wrote: > On maandag, maa 17, 2003, at 04:05 Europe/Brussels, Eric Grant wrote: > >> Based on my experience with some simple test programs (under OS X >> 10.2.3), >> and confirmed by my skimming of the sources (Revision 1.1.1.4 of >> malloc.c), >> it appears that realloc does not "free" memory (in any meaningful >> sense) > > realloc is a libc function, not a kernel function. As such, when you > free memory using realloc, this memory is put back in libc's memory > allocator pool for reuse for other malloc/realloc/calloc's. I never asserted that realloc was a kernel function (and I apologize for posting first on the wrong list). In any event, my point is precisely that the space is *NOT* reused for other malloc's, etc. The logical address space is gone, gone, gone. Do you disagree? Do you read the sources differently? Does a test program give you different results from mine on your machine(s)? Cheers, Eric _______________________________________________ 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

Disk Scheduling
by Joey Echeverria 18 Mar '03

18 Mar '03

I'm working on a project where we are using the Darwin kernel. I need to look into disk scheduling but was having a hard time traversing the source. I was hoping that you may be able to point me in the right direction. Thanks, Joey Echeverria Department of Electrical and Computer Engineering Carnegie Mellon University _______________________________________________ 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: string handling broken in dev_finddir(), bsd/miscfs/devfs/devfs_tree.c
by Quinn 17 Mar '03

17 Mar '03

At 19:55 +0100 17/3/03, Daniel Mack wrote: There is a bug in the string handling in dev_finddir(), bsd/miscfs/devfs/devfs_tree.c Please file a bug report so that the relevant engineering folks get to see this. <http://developer.apple.com/bugreporter/> S+E -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Technical Support * Networking, Communications, Hardware _______________________________________________ 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: Max number partitions per disk??
by Dan Markarian 17 Mar '03

17 Mar '03

Hi Wade, A specific partition scheme might have limits, but there is no system limit and has never been. The BSD disklabel is just another partition style in this big picture. The system has no dependence on a specific partition style, architecturally. Dan On Monday, March 17, 2003, at 04:38 PM, Wade Tregaskis wrote: >From /usr/include/sys/disklabel.h: /* * The absolute maximum number of disk partitions allowed. * This is the maximum value of MAXPARTITIONS for which 'struct disklabel' * is <= DEV_BSIZE bytes long. If MAXPARTITIONS is greater than this, beware. */ #define MAXMAXPARTITIONS 22 #if MAXPARTITIONS > MAXMAXPARTITIONS #warning beware: MAXPARTITIONS bigger than MAXMAXPARTITIONS #endif Wade Tregaskis _______________________________________________ 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: Max number partitions per disk??
by Wade Tregaskis 17 Mar '03

17 Mar '03

>From /usr/include/sys/disklabel.h: /* * The absolute maximum number of disk partitions allowed. * This is the maximum value of MAXPARTITIONS for which 'struct disklabel' * is <= DEV_BSIZE bytes long. If MAXPARTITIONS is greater than this, beware. */ #define MAXMAXPARTITIONS 22 #if MAXPARTITIONS > MAXMAXPARTITIONS #warning beware: MAXPARTITIONS bigger than MAXMAXPARTITIONS #endif Wade Tregaskis _______________________________________________ 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

string handling broken in dev_finddir(), bsd/miscfs/devfs/devfs_tree.c
by Daniel Mack 17 Mar '03

17 Mar '03

There is a bug in the string handling in dev_finddir(), bsd/miscfs/devfs/devfs_tree.c A string is not null terminated, see below. The bug takes affect when trying to generate a new entry with devfs_make_node() with a name containg a path component. The fix is simple, so I post it here: diff -u -r1.1.1.5 devfs_tree.c --- devfs_tree.c 26 Feb 2003 03:40:26 -0000 1.1.1.5 +++ devfs_tree.c 17 Mar 2003 16:30:57 -0000 @@ -335,7 +335,9 @@ scan++; strncpy(component, start, scan - start); - if (*scan == '/') + component[scan-start] = '\0'; + + if (*scan == '/') scan++; dirent_p = dev_findname(dirnode, component); Daniel _______________________________________________ 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: Different headers definitions between Kernel.framework and /usr/include/sys/socketvar.h
by Justin Walker 17 Mar '03

17 Mar '03

On Monday, Mar 17, 2003, at 02:18 US/Pacific, Stiphane Sudre wrote: On Friday, March 14, 2003, at 06:16 PM, Justin Walker wrote: On Friday, Mar 14, 2003, at 07:53 US/Pacific, Stiphane Sudre wrote: SS_ISDISCONNECTED is not declared in Kernel.framework. Is it correct? Incorrect. It is defined in <sys/socketvar.h>, which, on my system, is in both the /usr/include and the Kernel.framework trees. Ah, another crime made by the "Absolute Path bug" in PB. The Path to Kernel.framework was pointing to the 10.1 headers in my project. Side question: When you're first calling sendto/sendmsg, am I right to believe the socket state is at more SS_PRIV? I don't understand your question. The SS_PRIV bit is set at the time the socket is created, and at no other time. It is *only* set if the creating process is owned by root. The reason for this question was to find if it was possible to know from the socket state whether it was connected or not when a sendto/sendmsg was called for a connection oriented communication. The SS_CONNECTED bit tells you that. The SS_PRIV bit means the process is "privileged". >From the xnu source code, I was only able to see that the SS_PRIV flag was set in the sendto/sendit/socreate/etc. But some tests have shown that the SS_ISCONNECTED, SS_NBIO or SS_ASYNC can be set too. Correct. Regards, Justin -- Justin C. Walker, Curmudgeon-At-Large * Institute for General Semantics | Men are from Earth. | Women are from Earth. | Deal with it. *--------------------------------------*-------------------------------* _______________________________________________ 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: Max number partitions per disk??
by Dan Markarian 17 Mar '03

17 Mar '03

Hi Tim, There is no such limit. Dan On Friday, March 14, 2003, at 07:20 PM, Timothy Standing wrote: I remember reading on this list a year ago (or it could have been two) that there was a limit on the number of partitions the kernel supported on each disk. The post said that the limit was 16 partitions per disk. (I am assuming that the limit still exists.) I would prefer to keep users from running into this limit so I have a few questions about it: - Does anyone know whether this limit includes the partition which contains the partition map itself? - Does it include Apple_Free partitions? - Where is the limit implemented? Thanks in advance, Tim Standing SoftRAID, LLC _______________________________________________ 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: Different headers definitions between Kernel.framework and /usr/include/sys/socketvar.h
by Stéphane Sudre 17 Mar '03

17 Mar '03

On Friday, March 14, 2003, at 06:16 PM, Justin Walker wrote: On Friday, Mar 14, 2003, at 07:53 US/Pacific, Stiphane Sudre wrote: SS_ISDISCONNECTED is not declared in Kernel.framework. Is it correct? Incorrect. It is defined in <sys/socketvar.h>, which, on my system, is in both the /usr/include and the Kernel.framework trees. Ah, another crime made by the "Absolute Path bug" in PB. The Path to Kernel.framework was pointing to the 10.1 headers in my project. Side question: When you're first calling sendto/sendmsg, am I right to believe the socket state is at more SS_PRIV? I don't understand your question. The SS_PRIV bit is set at the time the socket is created, and at no other time. It is *only* set if the creating process is owned by root. The reason for this question was to find if it was possible to know from the socket state whether it was connected or not when a sendto/sendmsg was called for a connection oriented communication. >From the xnu source code, I was only able to see that the SS_PRIV flag was set in the sendto/sendit/socreate/etc. But some tests have shown that the SS_ISCONNECTED, SS_NBIO or SS_ASYNC can be set too. _______________________________________________ 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: realloc
by Eric Grant 17 Mar '03

17 Mar '03

On Sun, 16 Mar 2003 20:39:45 -0800, Justin Walker <justin(a)mac.com> wrote: > On Sunday, Mar 16, 2003, at 19:05 US/Pacific, Eric Grant wrote: > >> Based on my experience with some simple test programs (under OS X >> 10.2.3), >> and confirmed by my skimming of the sources (Revision 1.1.1.4 of >> malloc.c), >> it appears that realloc does not "free" memory (in any meaningful >> sense) >> when passed a size smaller than the size originally malloc'd. That is, >> despite the supposed downsizing of the object, none of the logical >> address >> space encompassed by the original object is available to satisfy >> subsequent >> allocation requests. >> >> For example: >> >> p1 = malloc (2 * 1024 * 1024 * 1024); // 2GB >> p2 = realloc (p1, 1); >> >> The result is that half the logical address space has basically >> disappeared >> (though it appears to be available to satisfy a request to realloc p2 >> to a >> size up to 2GB). >> >> To my mind, this result violates the standard (from man realloc): >> >>> The realloc() function changes the size of the previously allocated >>> mem- >>> ory referenced by ptr to size bytes. >> >> Moreover, it is difficult to see how this result is otherwise a >> "feature." >> >> Good? Bad? Ugly? > > None of the above? I'd say it was normal from my experience. Nothing > violates the description you quote above, at any rate. AFAIK, this is > the way most unices act for realloc, and in addition, address space > acquired in the process of malloc()ing space is never returned. It violates the description because to say that an allocated memory block referenced by ptr has a particular "size" is to say that the logical address space bounded by {ptr, ptr + size} is NOT available for later allocation and, conversely, that logical address space outside those bounds (and also outside all other allocated blocks) IS available for allocation. Surely, you don't mean to say that address space acquired in the process of malloc()ing space is not returned even when it is free()d? > This is really a topic for darwin-developers, not darwin-kernel. The > kernel just provides resources for user-mode operation, and it's > user-mode that dictates policy (at least for this issue). I've posted my message there. Thanks. Cheers, Eric Grant _______________________________________________ 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