January 2005 - Darwin-kernel

IOLog on serial port? Or to use kprintf?
by Parav Pandit 22 Jan '05

22 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Hi, Currently I see my IOLog() messages in to the /var/log/system.log file. Can I redirect it to my serial port? My Sytem is MAC OS X 10.3.7 on Power MAC G5. If yes, Do I have to set do nvram boot-args="debug=0x08" ? (DB_KPRT) I learned from Kernel Programming.pdf . I tried it, but I still get IOLog() messages in the same file. I hope I don't have to recompile the kernel for this. Or U can direct only kprintf() to serial port??? what should be serial port setting for MAC OS X G5? is it 57600 8N1 or 38400 8N1 ? G5 has that modem kind of RJ11 conector, not the serial port that we see on PC, so do I need to connect it via RJ11 connector to other host ? If I have to use kprintf() in my driver code, do I need to recompile the kernel, to see the messages on the console / serial port? I think need to recompile in DEBUG mode, right? I searched the xnu code, but I don't find any code of the kprintf in any of the c files. Where is the kprintf() definition??? Regards, Parav Pandit ________________________________________________________________________ Yahoo! India Matrimony: Find your life partner online Go to: http://yahoo.shaadi.com/india-matrimony _______________________________________________ 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: Interface layer level and packets
by Joshua Graessley 21 Jan '05

21 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com -josh On Jan 21, 2005, at 2:07 AM, Stephane Sudre wrote: On 20 janv. 05, at 18:35, Stephane Sudre wrote: This email sent to jgraessley(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 Be very careful when implementing packet fragment reassembly. It is possible for malicious hosts to fill up all of the buffers in the kernel if the fragment reassembly code doesn't set some limits. The darwin kernel contains some limitations. I'm seeing something I'm sure it's possible to deal with but haven't find the solution yet. I am at the interface level. I'm dealing with an incoming Ethernet "frame" (one buffer with a ether_header at is beginning). The buffer is within a mbuf chain. The IP header is indicating that the IP packet is 44 bytes long and the IP header is 20 bytes long (no options so). The problem I'm seeing is that the mbuf length is 46 byes after a mbuf pull up. So if I remove the ether_header length, I have only 32 bytes left. If I remove the IP headers, 12 bytes left. The issue is that it's a TCP/IP packet and so 14 bytes are at least necessary to get all the valuable TCP information (till the TCP flags). It's not one of the funny case described in RFC 1858 as the packet length includes the complete TCP header. The solution I have not found yet is: can I get the missing bytes somehow when a mbuf pullup can't get it? After rethinking about this, it's probably just standard packet fragmentations. On my way to re-invent the wheel to re-unit fragments... _______________________________________________ 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/jgraessley% 40apple.com smime.p7s

1 0

Re: Interface layer level and packets
by Stephane Sudre 21 Jan '05

21 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com On 20 janv. 05, at 18:35, Stephane Sudre wrote: _______________________________________________ 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 seeing something I'm sure it's possible to deal with but haven't find the solution yet. I am at the interface level. I'm dealing with an incoming Ethernet "frame" (one buffer with a ether_header at is beginning). The buffer is within a mbuf chain. The IP header is indicating that the IP packet is 44 bytes long and the IP header is 20 bytes long (no options so). The problem I'm seeing is that the mbuf length is 46 byes after a mbuf pull up. So if I remove the ether_header length, I have only 32 bytes left. If I remove the IP headers, 12 bytes left. The issue is that it's a TCP/IP packet and so 14 bytes are at least necessary to get all the valuable TCP information (till the TCP flags). It's not one of the funny case described in RFC 1858 as the packet length includes the complete TCP header. The solution I have not found yet is: can I get the missing bytes somehow when a mbuf pullup can't get it? After rethinking about this, it's probably just standard packet fragmentations. On my way to re-invent the wheel to re-unit fragments... This email sent to site_archiver(a)lists.apple.com

1 0

NKE Example
by Carl Smith 20 Jan '05

20 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Thread-index: AcT/QG38EyRIlMbJTDGZrA0M6j2Tvg== Thread-topic: NKE Example S+E -- Quinn "The Eskimo!" <http://www.apple.com/developer/> _______________________________________________ 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 At 8:33 +0100 2/10/04 , Mark Woollard wrote: I'm looking at writing an NKE and am looking for some sample code. The Apple document on the subject mentions TCPLogger as an example but I can't find the code online anywhere. Could someone point me to it or a similar sample? In general we try to discourage NKE development for platforms prior to Tiger because of the poor binary compatibility story on those platforms. Thus, we don't go out of our way to make the documentation and sample code easy to find. If you want to develop an NKE for a pre-Tiger platform, you should drop a line to <email@hidden> and we'll set you up with the latest.

1 0

Re: Consistent kernel panic on volume mount
by David Gatwood 20 Jan '05

20 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com I looked at this a bit earlier. This code seems to get called when replaying the journal. Would it work to simply mount the filesystem read-only (so as to avoid writes), and then to disable journaling using hfs.util? Or will the kernel insist on replaying the journal even with a read-only fs? Instead of mounting the volume, try just doing fsck_hfs -f -n /dev/disk#s# David _______________________________________________ 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.… On Wed, 19 Jan 2005, Carsten G Pedersen <Carsten.Pedersen(a)hasselblad.dk> wrote: Andrew Gallatin <gallatin(a)cs.duke.edu> wrote on 01/19/2005 10:07:48 PM: Unfortunately yes it does seem to be called when mounting read-only too - I get the same panic when doing a mount_hfs -o rdonly -j with # being the appropriate numbers for the volume in question. This will cause the journal to be ignored and disable any writes. See what errors it returns. If it doesn't show any serious errors, you might remove the -n. This -should- cause the journal (which sounds like it is corrupt) to be discarded, and at that point, you should be able to mount the disk. If that doesn't occur, something Really Weird(tm) is going on. That said, as someone else already mentioned, I'd be tempted to do a block-level backup of the drive to a file on a larger disk using dd before I ran fsck_hfs just to be safe. This email sent to site_archiver(a)lists.apple.com

1 0

Re: Consistent kernel panic on volume mount
by Carsten G Pedersen 20 Jan '05

20 Jan '05

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

Interface layer level and packets
by Stephane Sudre 20 Jan '05

20 Jan '05

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.… I'm seeing something I'm sure it's possible to deal with but haven't find the solution yet. I am at the interface level. I'm dealing with an incoming Ethernet "frame" (one buffer with a ether_header at is beginning). The buffer is within a mbuf chain. The IP header is indicating that the IP packet is 44 bytes long and the IP header is 20 bytes long (no options so). The problem I'm seeing is that the mbuf length is 46 byes after a mbuf pull up. So if I remove the ether_header length, I have only 32 bytes left. If I remove the IP headers, 12 bytes left. The issue is that it's a TCP/IP packet and so 14 bytes are at least necessary to get all the valuable TCP information (till the TCP flags). It's not one of the funny case described in RFC 1858 as the packet length includes the complete TCP header. The solution I have not found yet is: can I get the missing bytes somehow when a mbuf pullup can't get it? I was thinking that maybe I could paste re-assembled frame fragments (not IP fragments) in case it's the explanation but I have some troubles figuring out how it could be done since I just have the ether_header (which has no information related to this) and the ip header (and the IP packet is not fragmented and does not pertain to this level anyway). This email sent to site_archiver(a)lists.apple.com

1 0

Re: Consistent kernel panic on volume mount
by Karl Magdsick 19 Jan '05

19 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:references; b=Ypam6PNom4SP+sBe4328QwxWaB74v4TYTHLtPFQzUDQ47sGDwwWq+O/lb0PA/NqM2CtJn4CiXpxezygZ3aPCNjEjFUPdGxSM14o6/VWBQLHI5WHWzMZpSMUr4hdC6f3994Oo2zLvzuSOt1+LEyvy4a9HHoGGF0IjcOh2ofUWlxU= You can probably mount your HFS+ partitions from a LinuxPPC boot CD and copy data over the network or else onto a UFS partition. I found 3 or 4 Linux PPC boot CDs capable of mounting HFS+ partitions in read-only mode. Hopefully at least one Linux CD distro can mount UFS in read-write mode. SysRescCD.org provides one of the ISO images I found to work well for mounting HFS+ in read-only mode. Unfortunately, booting into single user mode from the Panther CD that came with my PowerBook didn't allow me to mount my HD's partition. Mount claimed the superblock was bad, but fsck said the partition was prefectly fine. Several Linux distros could mount my HD partition in read-only mode and open rc.boot with vi, but wouldn't let me edit rc.boot to remove a fatal bug I introduced. When I tried forcing a read-write mount mode, Linux silently ignored my request and mounted the partition read-only. Darwin still runs rc.boot in single-user mode, preventing booting if there is a problem with rc.boot. I ended up deciding to re-install Panther rather than wait for my 5th iso image of the night (OpenDarwin install CD) to burn. On a side note, are there any user-mode HFS+ (or other FS) drivers for xnu floating around? -Karl _______________________________________________ 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: Consistent kernel panic on volume mount
by Carsten G Pedersen 19 Jan '05

19 Jan '05

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 Andrew Gallatin <gallatin(a)cs.duke.edu> wrote on 01/19/2005 10:07:48 PM: > > I looked at this a bit earlier. This code seems to get called when > replaying the journal. Would it work to simply mount the filesystem > read-only (so as to avoid writes), and then to disable journaling > using hfs.util? Or will the kernel insist on replaying the journal > even with a read-only fs? > Unfortunately yes it does seem to be called when mounting read-only too - I get the same panic when doing a mount_hfs -o rdonly -j Carsten

1 0

Re: Consistent kernel panic on volume mount
by Andrew Gallatin 19 Jan '05

19 Jan '05

site_archiver(a)lists.apple.com Delivered-To: darwin-kernel(a)lists.apple.com Carsten G Pedersen writes: > > And yes Felix - I will take your advice and file a bug - just want to > gather some more information first. I'm seriously considering building a > version of the kernel that stays away from journaling altogether - of > course if someone else know about an easier way to avoid this code being > called I'd be most interested to hear about it. I looked at this a bit earlier. This code seems to get called when replaying the journal. Would it work to simply mount the filesystem read-only (so as to avoid writes), and then to disable journaling using hfs.util? Or will the kernel insist on replaying the journal even with a read-only fs? Drew _______________________________________________ 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