May 2005 - Darwin-kernel - lists.apple.com

Accessing to a file from a KEXT
by Pezeshgi, Shahriar 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcVPJ6YCkB1ysW1vTAqEcMyfC5oYHw== Thread-topic: Accessing to a file from a KEXT Hi there I have a problem with our USB DSL Modem driver under new OS X (Tiger). Basically we need to read a configuration file during driver startup ( I know that it's a BAD idea to access to file system from a KEXT but this is the way our driver has been developed originally under OS X 10.1). This driver is working properly under different version of OS X (10.1.x, 10.2.x and 10.3.x) but it's not working under Tiger, it's complaining about missing symbols like _vn_open, _vn_close and etc. I think new vnode api doesn't support old functions under Tiger (am I right?), so I looked at KPI->vnode documents but there is no descriptions for functions like vnode_open, vnode_close and etc. Where can I find more details about vnode api? Yours Shahriar _______________________________________________ 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: thread_funnel_set?
by Brian Bergstrand 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hey Sam! On May 2, 2005, at 4:43 AM, Sam Vaughan wrote: > The funnels are still there to let file systems that aren't re- > entrant continue to behave the way they did pre-Tiger. If you > don't set a VFS flag to indicate that you're a re-entrant file > system, the VFS/VOP interfaces will take the funnel before calling > your file system entry points. > > Should be a handy middle step when porting a 10.3 file system to > 10.4. Once you're up and running, turn on re-entrancy and watch > the deadlocks, data corruptions and panics roll in ;-) Hmmm, the way I've understood it is that when you set the non- reentrant flag, the VFS creates a mutex on the FS's behalf and then automatically acquires/releases it before any VFS entry point. There should be no need for a global kernel funnel, as the kernel protects it's own data structures. Time to look at the source! > > Sam > > >> Fore pre-Tiger, yes, thread funnels were basically giant locks. There >> was one for networking and one for everything else in the BSD kernel. >> I'm surprised that they are still in Tiger, as that is supposed to be >> one of the big changes in the Tiger kernel: no more funnels. I can >> only surmise that they are there for some kind of backward >> compatibility, but since all NKE and FS kext's have to be updated to >> the kpi's I don't know who that compatibility would be for. >> > Brian Bergstrand <http://www.classicalguitar.net/brian/> PGP Key ID: 0xB6C7B6A2 -----BEGIN PGP SIGNATURE----- Version: PGP Desktop 9.0.0 (Build 2001) iQA/AwUBQnY/TnnR2Fu2x7aiEQJPmACg5JLZyviKZvmLaDDCz1sSzLRj/mkAoI0l 8zGbfOH9wnGFeB9jMgYqKFJe =vgjV -----END PGP SIGNATURE----- _______________________________________________ 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

IODelay function
by Imran Kazi 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcVPIUxBujqCRWflRx+gIw30eHpppQ== Hi folks, I had written media filter scheme driver. I have implemented read/write routine in the driver. I am able to successfully write on the disk using this driver. Actually i have to hold the write routine for a small time interval say (few seconds),so that i call the read routine of my driver and read the data on the disk and then call the write routine. I am doing this to read the data on the disk and copy it to some other disk before my original write routine modifies the data on the disk. I am using IODelay routine to delay my driver write routine for a few seconds. But because of IODelay function my read routine also gets delay as i am specifying IODelay before calling read routine. Is it possible to delay only write call and not the read call, so that I call my driver read routine to get the data on the disk and then call write routine??? Thanks in advance. Imran _______________________________________________ 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

sbdrop - count not zero?
by Stephane Sudre 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com I've seen this strange kernel lof record in system.log on 10.4: "kernel[2] sbdrop - count not zero" What is it supposed to mean? Can it be related to a sudden stop of a TCP transfer/download? _______________________________________________ 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: thread_funnel_set?
by Sam Vaughan 02 May '05

02 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Sam Fore pre-Tiger, yes, thread funnels were basically giant locks. There was one for networking and one for everything else in the BSD kernel. I'm surprised that they are still in Tiger, as that is supposed to be one of the big changes in the Tiger kernel: no more funnels. I can only surmise that they are there for some kind of backward compatibility, but since all NKE and FS kext's have to be updated to the kpi's I don't know who that compatibility would be for. _______________________________________________ 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.… The funnels are still there to let file systems that aren't re- entrant continue to behave the way they did pre-Tiger. If you don't set a VFS flag to indicate that you're a re-entrant file system, the VFS/VOP interfaces will take the funnel before calling your file system entry points. Should be a handy middle step when porting a 10.3 file system to 10.4. Once you're up and running, turn on re-entrancy and watch the deadlocks, data corruptions and panics roll in ;-) This email sent to site_archiver(a)lists.apple.com

1 0

tracing the mount syscall
by Bijoy Thomas 01 May '05

01 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, Thanks in advance, Bijoy Thomas. _______________________________________________ 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.… I'm trying to write a new filesystem kext for the kernel. I just want to fill in the necessary vfsops to get the mount to succeed. After going through the mount source in vfs_syscalls.c, it seems that only the xxxfs_mount function is called. I wrote a dummy xxxfs_mount function to return 0, but my kernel panics when i run mount. I guess there is more to mount than just xxxfs_mount. I'm guessing that xxxfs_root and xxxfs_vget are called somewhere. Can anyone provide any info on what functions are called and when during a mount? This email sent to site_archiver(a)lists.apple.com

1 0

tracing the mount syscall
by Bijoy Thomas 01 May '05

01 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, Thanks in advance, Bijoy Thomas. _______________________________________________ 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.… I've been trying to write a filesystem kext and im trying to get the mount to work first. I want to fill in only the needed functions in the vfsops struct for the mount to succeed. After going through the mount function in vfs_syscalls.c, i see that only the fsXXX_mount function is called. I wrote a trivial fsXXX_mount function that returns 0, but my kernel goes into panic :( . I'm sure that there is more to mounting than just the fsXXX_mount function. I'm guessing that the fsXXX_vget and fsXXX_root functions are also called. Is there any documentation on the functions that are called during a mount? This email sent to site_archiver(a)lists.apple.com

1 0

Re: thread_funnel_set?
by Shawn Erickson 01 May '05

01 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Can anyone explain what the function thread_funnel_set() does? Is it some means of synchronization/locking mechanism to critical kernel data structures?? Fore pre-Tiger, yes, thread funnels were basically giant locks. There was one for networking and one for everything else in the BSD kernel. I'm surprised that they are still in Tiger, as that is supposed to be one of the big changes in the Tiger kernel: no more funnels. I can only surmise that they are there for some kind of backward compatibility, but since all NKE and FS kext's have to be updated to the kpi's I don't know who that compatibility would be for. The 10.3 and earlier style of kernel locking is outlined here... -Shawn _______________________________________________ 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 On Apr 30, 2005, at 9:34 PM, Brian Bergstrand wrote: On Apr 30, 2005, at 7:39 PM, Bijoy Thomas wrote: Note that compatibility libraries exist that you can link against (com.apple.kernel.*) if you are not KPI ready. I believe if you are using the compatibility libraries you cannot link against the KPI libraries and vice versa. <http://developer.apple.com/technotes/tn/ tn2028.html#MacOSXKernelThreading> It hasn't been updated for Tiger yet which as Brian noted has much finer grain locking in the BSD side of the kernel (IOKit already had good locking and concurrency support). smime.p7s

1 0

NKE sample code
by Vincent Pottier 01 May '05

01 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi list, I get TCPLog sample code from DTS a few month ago (Tiger version). I just try to rebuild the version I get and encounter a few fatal error, before searching to solve it, is there a new version already debugged? Thanks in advance. Vincent _______________________________________________ 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: thread_funnel_set?
by Brian Bergstrand 01 May '05

01 May '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On Apr 30, 2005, at 7:39 PM, Bijoy Thomas wrote: > Hi, > > Can anyone explain what the function thread_funnel_set() does? Is > it some means of synchronization/locking mechanism to critical > kernel data structures?? > > Regards, > Bijoy Thomas. > Fore pre-Tiger, yes, thread funnels were basically giant locks. There was one for networking and one for everything else in the BSD kernel. I'm surprised that they are still in Tiger, as that is supposed to be one of the big changes in the Tiger kernel: no more funnels. I can only surmise that they are there for some kind of backward compatibility, but since all NKE and FS kext's have to be updated to the kpi's I don't know who that compatibility would be for. Brian Bergstrand <http://www.classicalguitar.net/brian/> PGP Key ID: 0xB6C7B6A2 -----BEGIN PGP SIGNATURE----- Version: PGP Desktop 9.0.0 (Build 2001) iQA/AwUBQnRcc3nR2Fu2x7aiEQJQNwCdFhM/TyJXHVp4MlI/Iv97GDYBBTEAoKSK lQ7fqoyJoKqKvO0vIU+NKxFh =ekk4 -----END PGP SIGNATURE----- _______________________________________________ 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