Thread: Re: [Etherboot-developers] Q: TODO list for 5.2
Brought to you by:
marty_connor,
stefanhajnoczi
Menu
▾
▴
etherboot-developers
|
From: <ke...@et...> - 2003-07-11 08:33:43
|
>Do we have a TODO list of what needs to happen before 5.2 is released? Mea culpa, I should really make this list. Marty and others, please post your lists. For me: pcnet32 support tlan support (if possible) UNDI (ready when you are Michael) documentation I'd like to get 5.2 out in time for LWE SF. |
|
From: <ke...@et...> - 2003-07-11 14:16:08
|
>The cvs includes my tlan attempt that is able to poll correctly and >transmit once. I suspect that a review of the documentation would point >out the issue. I was not able to get past the autonegotiation stage on my OC2336 even though the lights on the hub seem to indicate the hub is happy with the speed. You have some recursive calls in TLan_FinishAutoNeg, those can't be right. You don't need ISA support in the driver, it's a PCI only card. I've reduced the number of Rx and Tx buffers and checked the changes in. |
|
From: Timothy L. <tl...@ro...> - 2003-07-14 01:07:13
|
> I was not able to get past the autonegotiation stage on my OC2336 even > though the lights on the hub seem to indicate the hub is happy with the > speed. You have some recursive calls in TLan_FinishAutoNeg, those can't > be right. The Linux driver does some strange things in the auto-negotiation. I will review my changes. I may be able to get my hands on another tlan car to test with also. > You don't need ISA support in the driver, it's a PCI only card. I've > reduced the number of Rx and Tx buffers and checked the changes in. Yeah, I left it in because the Linux driver mentioned ISA versions. Obviously it should have been taken out. I am not sure whether the Tx buffer can be set to 1. The card requires the forward pointer of the ring to point to the next buffer. If it sees a zero it ends the channel. (Which could be Ok, I have tried to make it do that to no avail.) If you point the forward pointer to itself it keeps transmitting the same thing... Thanks Tim --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.497 / Virus Database: 296 - Release Date: 7/4/2003 |
|
From: Timothy L. <tl...@ro...> - 2003-07-14 13:50:22
|
> > I was not able to get past the autonegotiation stage on my OC2336 even > > though the lights on the hub seem to indicate the hub is happy with > the > > speed. You have some recursive calls in TLan_FinishAutoNeg, those > can't > > be right. > > The Linux driver does some strange things in the auto-negotiation. I > will review my changes. I may be able to get my hands on another tlan > car to test with also. If I remember correctly and I think I do. The Linux driver sets up an interrupt based timer to allow it to restart parts of the auto negotiation if it need more time. Since eb lacks an interrupt handler I improvised by calling the function that the linux driver would have when the interrupt kicked in. If you can think of a better way let me know. Also, where exactly in the negotiation does it fail? Can you attach a serial debug? Also, as far as I can tell using one transmit buffer with the tlan is not possible. I believe that the minimum is two because unlike some cards where you simply write the address of the ring again it depends on the forward pointer of the previous tx buffer pointing to the next tx buffer. There may be a way in the doc to prevent it from continuing to transmit the wrapped ring I am not aware of it... Tim --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.497 / Virus Database: 296 - Release Date: 7/4/2003 |
|
From: <ke...@et...> - 2003-07-11 14:52:15
|
>I started pcnet32 a while back, but without a hardware nic, it became >difficult. For some reason I hit a size issue in 5.0 and I did not seem >to be able to get 5.1 to boot via vmware. I would add fixing vmware >support to the list, but I don't know where to start. I am however >looking to purchase a pcnet based NIC if I can find one. They seem to >be fairly rare... You didn't leave the number of Rx and Tx buffers at the Linux default did you? That would blow up the footprint. I usually bring the number of Rx buffers down to 2 or 4 and the Tx buffer to 1 for Etherboot. The protocols used are stop and wait anyway, and don't need to transmit back to back. |
|
From: Timothy L. <tl...@ro...> - 2003-07-14 01:08:12
|
> You didn't leave the number of Rx and Tx buffers at the Linux default > did you? That would blow up the footprint. I usually bring the number of > Rx buffers down to 2 or 4 and the Tx buffer to 1 for Etherboot. The > protocols used are stop and wait anyway, and don't need to transmit back > to back. You may be right, I will have to check that... --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.497 / Virus Database: 296 - Release Date: 7/4/2003 |
|
From: Timothy L. <tl...@ro...> - 2003-07-11 10:07:12
Attachments:
minicom.cap
|
> pcnet32 support > tlan support (if possible) > UNDI (ready when you are Michael) > documentation I started pcnet32 a while back, but without a hardware nic, it became difficult. For some reason I hit a size issue in 5.0 and I did not seem to be able to get 5.1 to boot via vmware. I would add fixing vmware support to the list, but I don't know where to start. I am however looking to purchase a pcnet based NIC if I can find one. They seem to be fairly rare... As for the tlan, I am attaching a minicom.cap file with hex dumps of what the poll received. It looks fine to me, but I am not sure. I suspect that there is something very small in the transmit function that I am overlooking. I would really like to get it in 5.2 but I will certainly back-port it when it works. I also started some detailed linux driver to Etherboot driver porting documentation when I started the pcnet driver. I should be able to finish that up after the tlan. Since it is not really version specific it should not be a dependency for 5.2. Also, since we now have the majority of the linux drivers supported there may be limited value in proceeding... Tim --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.497 / Virus Database: 296 - Release Date: 7/4/2003 |
|
From: Anselm M. H. <an...@ho...> - 2003-07-12 15:03:45
|
Hello Timothy, > I also started some detailed linux driver to Etherboot driver porting > documentation when I started the pcnet driver. I should be able to > finish that up after the tlan. Since it is not really version specific > it should not be a dependency for 5.2. Also, since we now have the > majority of the linux drivers supported there may be limited value in > proceeding... One of my long-term-projects ;-) is to port the Realtek 8169 driver. It seems similar to 8139, with some not-so-nice (and not well documented, I think) changes. It is a gigabit NIC, and having my workstation boot an LTSP for the occasional mail check instead of a full-blown Debian would be way cool. My time is limited though, as the usual exams are approaching. Perhaps your docs would give me at least a good starting point. Could you mail them (incomplete if they are, no problem)? Regards, Anselm |
|
From: <ebi...@ln...> - 2003-07-11 10:21:32
|
ke...@et... (Ken Yap) writes: > >Do we have a TODO list of what needs to happen before 5.2 is released? > > Mea culpa, I should really make this list. Marty and others, please post > your lists. For me: > > pcnet32 support > tlan support (if possible) > UNDI (ready when you are Michael) > documentation > > I'd like to get 5.2 out in time for LWE SF. Multicast reception support in most of the drivers. Besides being cool on large clusters. We will probably need it if we ever support IPv6, as a lot of broadcast traffic has been replaced with multicast traffic. Eric |
|
From: Marty C. <md...@et...> - 2003-07-11 10:54:35
|
On Friday, July 11, 2003, at 04:31 AM, Ken Yap wrote: >> Do we have a TODO list of what needs to happen before 5.2 is released? > > Mea culpa, I should really make this list. Marty and others, please > post > your lists. For me: > > pcnet32 support > tlan support (if possible) > UNDI (ready when you are Michael) > documentation > > I'd like to get 5.2 out in time for LWE SF. I think 5.2 for LWE might be a little ambitious, though possibly doable. My ToDo list is mainly a testing one. I'd like to create a matrix something like: Etherboot Driver .zrom .zdsk .zpxe .zlilo tulip OK OK OK Failed rtl8139 OK OK OK OK 3C905B OK OK OK OK You get the idea. I'd just like to have confidence before we release that the basic functionality and traditional robustness is there. I have a fair number of cards here, and a testing setup, ROM burner, etc. So I can help with this. Of course, many people pitching in would be a real help! 5.1.x has some fairly drastic changes (for the better!) to the code, and so I'd like to take a little extra time testing before releasing as production quality. It's been a busy summer for me, and it will likely remain so, but I'd also really like get 5.2 out, but not at cost of stability. Of course, as usual, until we release, we won't get the quality and quantity of bug messages that we so much enjoy ;) Then there's the LTSP issue. I've had a message on the rom-o-matic.net site, warning LTSP users not to use 5.1.x because of incompatible changes. Jim, if you're listening, could you suggest alternative wording, now that the new kernel package is out? I'm sure we'd get a lot of LTSP users testing 5.1.x is we said something like: LTSP users must upgrade to the latest ltsp-kernel package (currently ltsp_kernel-3.0.10) available at: http://www.ltsp.org/instructions-3.0.html in order to use Etherboot 5.1.x Can you let me know if this is correct, or suggest an alternative. We value feedback from LTSP users, and want to make sure we keep them happy and don't disappoint them with booting problems. Other than that, I agree with Ken's list. I'd love to take a crack at the TLAN driver. I have a card here somewhere, and it would be fun. It would take two solid days of work to do it, though, and I'm not sure I can commit that with my current workload. So that's my view of things. I'm excited by the possibility of getting 5.2 out. If everyone pitches in with some testing and we do a bug hunt, I'll feel better about putting it out there. Marty -- Try: http://rom-o-matic.net/ to make Etherboot images instantly. Name: Marty Connor US Mail: Entity Cyber, Inc.; P.O. Box 391827; Cambridge, MA 02139; USA Voice: (617) 491-6935; Fax: (617) 491-7046 Email: md...@et... Web: http://www.etherboot.org/ |
|
From: Timothy L. <tl...@ro...> - 2003-07-11 11:02:42
|
> Other than that, I agree with Ken's list. I'd love to take a crack at > the TLAN driver. I have a card here somewhere, and it would be fun. It > would take two solid days of work to do it, though, and I'm not sure I > can commit that with my current workload. The cvs includes my tlan attempt that is able to poll correctly and transmit once. I suspect that a review of the documentation would point out the issue. Tim --- Outgoing mail is certified Virus Free. Checked by AVG anti-virus system (http://www.grisoft.com). Version: 6.0.497 / Virus Database: 296 - Release Date: 7/4/2003 |
|
From: <ja...@Mc...> - 2003-07-11 14:41:11
|
On Fri, 11 Jul 2003, Marty Connor wrote: > > Then there's the LTSP issue. I've had a message on the rom-o-matic.net > site, warning LTSP users not to use 5.1.x because of incompatible > changes. Jim, if you're listening, could you suggest alternative > wording, now that the new kernel package is out? I'm sure we'd get a > lot of LTSP users testing 5.1.x is we said something like: > > LTSP users must upgrade to the latest ltsp-kernel package > (currently ltsp_kernel-3.0.10) available at: > > http://www.ltsp.org/instructions-3.0.html > > in order to use Etherboot 5.1.x > > Can you let me know if this is correct, or suggest an alternative. We > value feedback from LTSP users, and want to make sure we keep them > happy and don't disappoint them with booting problems. Marty, I think you've said it perfectly. Also, looking forward to another great Linux World. Make sure you keep Wednesday evening free for the traditional LTSP dinner. Jim. |
