April 2011 - Darwin-kernel - lists.apple.com

Re: Bug in NFS
by Shantonu Sen 08 Apr '11

08 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Bug in NFS
by Feng Shuo 08 Apr '11

08 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Dkim-signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=gamma; h=domainkey-signature:mime-version:from:date:message-id:subject:to:cc :content-type; bh=0y8dOAIWPw1QLZAsuskQFfSNgMfwpD7hpXfjVqSUTUg=; b=bq2dDE9msz2dz89bBo0pfUOobH4Va3wMS+tLhOm2m9f+aE6e4QVs/XfTBllsW9CFJn LhXPkZULruBQSOAPX/Zg+q3YkaoIh2xScPdK2gCYFFWkH+Y/zuI400uZofxx5equFk1X QQKMk+YUgBBhNwFDObgn8dIYSKEyuU1Hfnj8c= Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:from:date:message-id:subject:to:cc:content-type; b=rFkftxlCBl3rmUeivo6BVW/EZScbrK00vjb08FglVFIN6a9ZFaCx5UiudpT6d42Fi3 QWBYfAtvyOMqDItO7AX2+ATXqtAGhN5i9Id3ErJvIaEeFulvDuGtrHwtQxnvZGU9UzwZ CTOVS04eHJLnRHGPEAvrwrW5hTl6JzBB86aI8= _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com nfs_v_name.patch

1 0

Re: Boot flags
by Justin C. Walker 07 Apr '11

07 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, Paul, On Apr 7, 2011, at 11:37 , Paul Nelson wrote: > Don't think the srv flag is documented there though. Unfortunately, this flag sets the global "srv" which is an unfortunate name for searching on... > > places to look at: Thanks for the hint. I'll check it out. One 'srv' chang that makes me a little nervous is this one: (Before/After) < net.inet.ip.rtexpire: 710 --- > net.inet.ip.rtexpire: 3600 End systems shouldn't latch onto routes for that long. But that's for another list... Justin -- Justin C. Walker, Curmudgeon at Large Institute for the Absorption of Federal Funds -- Democracy is two wolves and a lamb voting on what to have for lunch. Liberty is a well-armed lamb contesting the vote. _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Classify VNODE SCOPE actions for file operations
by Ken Hornstein 07 Apr '11

07 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com >I think this discussion is probably moot anyway because even if you get >a name, you can't depend on using it to find a vnode again (consider >renaming, deleting, etc). >You could use it for logging purposes I suppose. Well, if you wanted to make a kauth decision based on the name (which I admit is a bit weird to me, but hey, whatever), then you could use it at that time. The original poster wasn't exactly specific on what they were interested in doing. >True for KAUTH_VNODE_ADD_FILE - you have different combos of vnodes >depending on the real operation. The vfs may not be trying to create a >file at all - see access1, rename. I guess it really should be labelled "ADD_FILE_TO_DIRECTORY". >I don't know about all filesystems, but HFS, Apple's SMB , Thursby's >CIFS vfs return it. >NFS doesn't, but I would bet that AFP does. I know va_name is used by >coreservicesd, and is how FSGetCatalogInfo calls get the name. Sure, but getattrlist_internal() will simply use vnode_getname() if the underlying VNOP_GETATTR() doesn't return anything for va_name. My point is that you can't rely on it; I guess to be portable, you should try seeing if the filesystem returns va_name and if it doesn't then you should fall back on vnode_getname(). Although from private conversatons I've had with Apple people, they hinted that vnode_getname() was actually pretty reliable. --Ken _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Classify VNODE SCOPE actions for file operations
by Paul Nelson 07 Apr '11

07 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com I think this discussion is probably moot anyway because even if you get a name, you can't depend on using it to find a vnode again (consider renaming, deleting, etc). You could use it for logging purposes I suppose. On Apr 7, 2011, at 1:01 PM, Ken Hornstein wrote: > I don't think that's going to work in this case. As I understand it, the > original poster asked about getting the filename at authorization (via > kauth) time. Unfortunately, you're not given a target vnode to the kauth > function because the vnode hasn't been created yet. What would be > ideal would be for the desired filename to be passed down to the kauth > routine, but that isn't done. True for KAUTH_VNODE_ADD_FILE - you have different combos of vnodes depending on the real operation. The vfs may not be trying to create a file at all - see access1, rename. For KAUTH_VNODE_READ_DATA you do have a vnode for the item. >> It is not that hard to use getattrlist inside a kernel. This is the >> best way to get the name of a vnode. >> Just fill out the vnop_getattr_args and call VNOP_GETATTRLIST. For >> example: >> [...] > > You mean VNOP_GETATTR(), right? I will note that from what I can see, Yep. > only hfs actually returns va_name on MacOS X (okay, it's pretty likely > you're going to be using hfs at least for local filesystems, but I guess > my point is that it's not a guarantee that it will work). > > --Ken I don't know about all filesystems, but HFS, Apple's SMB , Thursby's CIFS vfs return it. NFS doesn't, but I would bet that AFP does. I know va_name is used by coreservicesd, and is how FSGetCatalogInfo calls get the name. Paul Nelson Thursby Software Systems, Inc. _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Boot flags
by Paul Nelson 07 Apr '11

07 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Don't think the srv flag is documented there though. Unfortunately, this flag sets the global "srv" which is an unfortunate name for searching on... places to look at: paul2:xnu-1504.9.37 nelson$ grep -wrI srv . // affects the mbuf_poolsz ./bsd/dev/unix_startup.c:int srv = 0; /* Flag indicates a server boot when set */ ./bsd/dev/unix_startup.c: //(void) PE_parse_boot_argn("srv", &srv, sizeof (srv)); ./bsd/dev/unix_startup.c: if (ncl || srv) ./bsd/dev/unix_startup.c: nmbclusters = mbuf_default_ncl(srv, sane_size); ./bsd/kern/bsd_init.c:extern int srv; // tunes the mbuf pool size for 64 bit kernels according to these tables: #if defined(__LP64__) typedef struct ncl_tbl { uint64_t nt_maxmem; /* memory (sane) size */ uint32_t nt_mbpool; /* mbuf pool size */ } ncl_tbl_t; /* Non-server */ static ncl_tbl_t ncl_table[] = { { (1ULL << GBSHIFT) /* 1 GB */, (64 << MBSHIFT) /* 64 MB */ }, { (1ULL << (GBSHIFT + 3)) /* 8 GB */, (96 << MBSHIFT) /* 96 MB */ }, { (1ULL << (GBSHIFT + 4)) /* 16 GB */, (128 << MBSHIFT) /* 128 MB */ }, { 0, 0 } }; /* Server */ static ncl_tbl_t ncl_table_srv[] = { { (1ULL << GBSHIFT) /* 1 GB */, (96 << MBSHIFT) /* 96 MB */ }, { (1ULL << (GBSHIFT + 2)) /* 4 GB */, (128 << MBSHIFT) /* 128 MB */ }, { (1ULL << (GBSHIFT + 3)) /* 8 GB */, (160 << MBSHIFT) /* 160 MB */ }, { (1ULL << (GBSHIFT + 4)) /* 16 GB */, (192 << MBSHIFT) /* 192 MB */ }, { (1ULL << (GBSHIFT + 5)) /* 32 GB */, (256 << MBSHIFT) /* 256 MB */ }, { (1ULL << (GBSHIFT + 6)) /* 64 GB */, (384 << MBSHIFT) /* 384 MB */ }, { 0, 0 } }; #endif /* __LP64__ */ ./bsd/kern/uipc_mbuf.c:mbuf_default_ncl(int srv, uint64_t mem) ./bsd/kern/uipc_mbuf.c:#pragma unused(srv) ./bsd/kern/uipc_mbuf.c: ncl_tbl_t *tbl = (srv ? ncl_table_srv : ncl_table); // tunes task_max, task_threadmax, thread_max for 64 bit machines with more than 16G ram ./osfmk/kern/startup.c:extern int srv; ./osfmk/kern/startup.c: if (PE_parse_boot_argn("srv", &srv, sizeof (srv))) { ./osfmk/kern/startup.c: srv = 1; ./osfmk/kern/startup.c: if ((srv != 0) && ((uint64_t)sane_size >= (uint64_t)(16 * 1024 * 1024 *1024ULL))) { On Apr 7, 2011, at 1:15 PM, Justin C. Walker wrote: > > On Apr 7, 2011, at 10:58 , Shantonu Sen wrote: > >> There's a com.apple.Boot.plist(5) man page. > > A man page?? Whoda thunk? > >> You'd have to read the source code. grep for PE_parse and then srv. > > Cool. I can do that. > > Thanks!! > > Justin > > -- > Justin C. Walker, Curmudgeon at Large > Director > Institute for the Enhancement of the Director's Income > ----------- > Nobody knows the trouble I've been > ----------- > > > > _______________________________________________ > 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: > http://lists.apple.com/mailman/options/darwin-kernel/nelson%40thursby.com > > This email sent to nelson(a)thursby.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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Boot flags
by Justin C. Walker 07 Apr '11

07 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On Apr 7, 2011, at 10:58 , Shantonu Sen wrote: > There's a com.apple.Boot.plist(5) man page. A man page?? Whoda thunk? > You'd have to read the source code. grep for PE_parse and then srv. Cool. I can do that. Thanks!! Justin -- Justin C. Walker, Curmudgeon at Large Director Institute for the Enhancement of the Director's Income ----------- Nobody knows the trouble I've been ----------- _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Classify VNODE SCOPE actions for file operations
by Ken Hornstein 07 Apr '11

07 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com >> It would seem the answer is ... no. At least not via kauth. It looks like >> it would require some changes internally; right now you would need to get >> passed in the struct nameidata down, and vnode_authorize only takes two >> vnodes as arguments. > >You could use vn_getpath (see vnode.h) but it does not ask the >filesystem for names, so it can't handle hard link nodes, and might >return an error. I don't think that's going to work in this case. As I understand it, the original poster asked about getting the filename at authorization (via kauth) time. Unfortunately, you're not given a target vnode to the kauth function because the vnode hasn't been created yet. What would be ideal would be for the desired filename to be passed down to the kauth routine, but that isn't done. >It is not that hard to use getattrlist inside a kernel. This is the >best way to get the name of a vnode. >Just fill out the vnop_getattr_args and call VNOP_GETATTRLIST. For >example: >[...] You mean VNOP_GETATTR(), right? I will note that from what I can see, only hfs actually returns va_name on MacOS X (okay, it's pretty likely you're going to be using hfs at least for local filesystems, but I guess my point is that it's not a guarantee that it will work). --Ken _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Re: Boot flags
by Shantonu Sen 07 Apr '11

07 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com There's a com.apple.Boot.plist(5) man page. You'd have to read the source code. grep for PE_parse and then srv. Shantonu Sen ssen(a)apple.com On Apr 7, 2011, at 10:41 AM, Justin C. Walker wrote: > Hi, all, > > A quick (one hopes) question: is the content of boot-args (currently) documented anywhere? I know one way that should work (to boost the per-uid process count), but I don't know what other side-effects setting "srv=1" will have. > > I checked the Apple dev site, and searched on the technotes page (0 hits for "boot"??). > > Thanks for help, as always. > > Justin > > -- > Justin C. Walker, Curmudgeon at Large > Institute for the Absorption of Federal Funds > ----------- > Like the ski resort full of girls hunting for husbands > and husbands hunting for girls, the situation is not > as symmetrical as it might seem. > - Alan MacKay > -- > > _______________________________________________ > 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: > http://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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0

Boot flags
by Justin C. Walker 07 Apr '11

07 Apr '11

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, all, A quick (one hopes) question: is the content of boot-args (currently) documented anywhere? I know one way that should work (to boost the per-uid process count), but I don't know what other side-effects setting "srv=1" will have. I checked the Apple dev site, and searched on the technotes page (0 hits for "boot"??). Thanks for help, as always. Justin -- Justin C. Walker, Curmudgeon at Large Institute for the Absorption of Federal Funds ----------- Like the ski resort full of girls hunting for husbands and husbands hunting for girls, the situation is not as symmetrical as it might seem. - Alan MacKay -- _______________________________________________ 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: http://lists.apple.com/mailman/options/darwin-kernel/site_archiver%40lists.… This email sent to site_archiver(a)lists.apple.com

1 0