|
|
|
|
||||||
| comp.protocols.tcp-ip TCP and IP network protocols. |
![]() |
|
|
LinkBack | Outils de la discussion |
|
|
#1 |
|
Messages: n/a
Hébergeur: |
Hi
Please point me to a 16-bit GPL'ed open source TCP/IP stack. I need to use during the 16-bit real mode during pre-boot environment. Does any one know if there are any current linux(uClinux, eCOS , RTlinux etc) tcp/ip stacks which can be used in 16-bit real mode. Anyone's prior experience with using a 16-bit tcp/ip stack. thanks for your -vikas |
|
|
|
#2 |
|
Messages: n/a
Hébergeur: |
What do you mean by 16-bit TCP?
TCP doesn't care about native machine bit sizes and instruction sizes. Some code may be written that assumes 32-bit integer math is cheap. But that's usually just the checksum routine -- easy to change. Craig "vik.reck@gmail.com" <vik.reck@gmail.com> writes: >Hi > Please point me to a 16-bit GPL'ed open source TCP/IP stack. I >need to use during the 16-bit real mode during pre-boot environment. >Does any one know if there are any current linux(uClinux, eCOS , >RTlinux etc) tcp/ip stacks which can be used in 16-bit real mode. >Anyone's prior experience with using a 16-bit tcp/ip stack. >thanks for your >-vikas |
|
|
|
#3 |
|
Messages: n/a
Hébergeur: |
In article <dv7s65$dvt$1@pcls4.std.com>,
Craig Partridge <craigp@std.com> top-posted, now corrected: >"vik.reck@gmail.com" <vik.reck@gmail.com> writes: >>Does any one know if there are any current linux(uClinux, eCOS , >>RTlinux etc) tcp/ip stacks which can be used in 16-bit real mode. >What do you mean by 16-bit TCP? >TCP doesn't care about native machine bit sizes and instruction sizes. On the other hand if you have a processor with a 16-bit non-memory-mapped address space, then you are going to have to be careful about window sizes and fragmentation reassembly and other items that assume that you have 64K objects readily available. |
|
|
|
#4 |
|
Messages: n/a
Hébergeur: |
roberson@hushmail.com (Walter Roberson) writes:
>>TCP doesn't care about native machine bit sizes and instruction sizes. >On the other hand if you have a processor with a 16-bit >non-memory-mapped address space, then you are going to have to be >careful about window sizes and fragmentation reassembly and other >items that assume that you have 64K objects readily available. TCP doesn't assume that. After all, some of the first implementations were done on 16-bit native machines. I just did a quick check (so I may have missed something) but the window size is 16 bits, as is the urgent pointer. Sequence numbers are 32-bits so you need to do some 32-bit arithmetic, but not much and it isn't very hard. The maximum packet size is 16 bits, so you're safe there. Certainly you don't, say, want to negotiate TCP big windows or extended sequence numbers -- but in a 16-bit machine, you're hardly ready to accept that kind of firehose of data anyway. Craig |
|
|
|
#5 |
|
Messages: n/a
Hébergeur: |
> Please point me to a 16-bit GPL'ed open source TCP/IP stack. I
>need to use during the 16-bit real mode during pre-boot environment. IMHO your best bet is to look for "embeddable TCP" stacks. There are a few open source projects in this area and they should be easily portable to 16 bit real mode PC envireonement. LwIP comes to mind, but there are others. Once you find one, links found on such a page most likely will lead you to the others. HTH Markus |
|
|
|
#6 |
|
Messages: n/a
Hébergeur: |
In article <dv92i6$otg$1@pcls4.std.com>,
Craig Partridge <craigp@std.com> wrote: >>items that assume that you have 64K objects readily available. > >TCP doesn't assume that. After all, some of the first implementations >were done on 16-bit native machines. The trouble is not in the assumptions of protocol but the assumptions of the implementors. Recall the recent talk here about TCP implementations that go (or went?) crazy when presented with "negative" port numbers and window sizes. I agree that a new TCP implementation for a 16-bit system would not be hard write. Such an extremely tiny system (by modern standards) should omit many modern elaborations from header predicting to large windows to varys buffering systems such as "cluster mbufs." The problem is that a very tiny (by today's standards) 16-bit system is not likely to be written from scratch instead of created by trimming and recompiling existing code that assumes that int's contain at least 32 bits. Vernon Schryver vjs@rhyolite.com |
|
|
|
#7 |
|
Messages: n/a
Hébergeur: |
In article <1142277506.265880.50490@j52g2000cwj.googlegroups. com>,
vik.reck@gmail.com <vik.reck@gmail.com> wrote: >Hi > Please point me to a 16-bit GPL'ed open source TCP/IP stack. I >need to use during the 16-bit real mode during pre-boot environment. > >Does any one know if there are any current linux(uClinux, eCOS , >RTlinux etc) tcp/ip stacks which can be used in 16-bit real mode. > >Anyone's prior experience with using a 16-bit tcp/ip stack. > >thanks for your >-vikas > Open Source DOS probably has a real mode IP stack. I don't know what the usage license is. http://www.freedos.org/ -- a d y k e s @ p a n i x . c o m Don't blame me. I voted for Gore. |
|
|
|
#8 |
|
Messages: n/a
Hébergeur: |
|
|
|
|
#9 |
|
Messages: n/a
Hébergeur: |
>very tiny (by today's standards) 16-bit system is not likely to be
>written from scratch instead of created by trimming and recompiling >existing code that assumes that int's contain at least 32 bits. That's why I sugested to look in the embedded area. There, all stacks known to me (including the one I wrote) were written from scratch. Markus |
|
![]() |
| Outils de la discussion | |
|
|