12:14:26 | Sammy | hello !! |
14:47:26 | erikm | hi |
14:49:21 | Russ | morning |
16:50:56 | Russ|werk | erikm: do you have that slightly fixed up version of the cs89x0 driver I sent out? |
16:53:15 | erikm | Russ|werk: yes |
16:53:23 | erikm | Russ|werk: I also hacked on it a bit |
16:53:35 | erikm | the IRQ 0 stuff was still not right |
16:53:56 | Russ|werk | IRQ can be signed and negative for no irq |
16:54:42 | Russ|werk | feel free to rip out teh memory mode stuff, as I don't think memory mode would actually be any faster |
16:54:58 | Russ|werk | the timings are almost the same for the chip |
16:55:15 | Russ|werk | so register accesses would just happen in half the time |
16:55:49 | erikm | the only way it would be faster is when you copy the packets in and out of the buffer |
16:56:04 | Russ|werk | why though? |
16:56:19 | Russ|werk | in IO mode, its doing successive 16 bit reads to the same location |
16:56:36 | Russ|werk | in MEM mode, its doing successive 16 bit reads to an incrementing location |
16:57:09 | erikm | ehm, yes. no real difference |
16:58:54 | Russ|werk | the only way to get more speed would be to hook up an SRAM buffer that emulates DMA |
16:59:12 | Russ|werk | and the CPU can read from at much higher speeds than the CS8900A |
17:02:19 | Russ|werk | anyway, wookie was looking for a working version |
17:04:29 | erikm | yes, I just just read the LART list |
17:04:41 | erikm | I'll see if I can get a hacked LART version out. |
17:04:54 | erikm | maybe even a complete patch set for 2.4.6-rmk2-np1 |
17:06:59 | Russ|werk | the driver works in the kernel |
17:07:21 | Russ|werk | it needs ether=... ether=... ether=... to enable multiple chips |
17:10:27 | Russ|werk | I guess the bigest things would be multiple chip support, loads when there is no link, and mii ioctl emulation |
17:10:52 | Russ|werk | I think it should return 0x0000, 0x0000 for vendor/device though... |
23:26:08 | erikm | votes for rmk going to -ac patches |