In article <Feb.3.23.55.22.1992.9885@dumas.rutgers.edu> hedrick@dumas.rutgers.edu (Charles Hedrick) writes:This is how 0.13 will do things: my current driver just sets a bit (this is part of the better timer-routines I wrote just for things like this), and the characters are then processed under the timer interrupt. This way, we still get instant echoing (well, it waits for the timer-intr, but that happens fairly often: 100 Hz), but we don't get the copy_to_cooked() function call several thousands of times a second with speeds > 19200. The current linux still has a limit of 2kB of buffer, so if the characters aren't read out of the buffer fast enough, you still lose characters due to buffer overrun. Also, linux doesn't support the 16550 chip in buffered mode: I don't have the hardware. Linus
| Len Brown | [PATCH 05/85] ACPI: Add "acpi.power_nocheck=1" to disable power state check in pow... |
| Andi Kleen | [PATCH] [2/50] x86_64: use core id bits for apicid_to_node initialization |
| Greg KH | [GIT PATCH] driver core patches against 2.6.24 |
| Andrew Morton | 2.6.23-rc1-mm1 |
git: | |
| Gregory Haskins | [RFC PATCH 06/17] ioq: Add basic definitions for a shared-memory, lockless queue |
| Arjan van de Ven | Re: [GIT]: Networking |
| Jarek Poplawski | [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| David Miller | Re: mac80211 truesize bugs |
