From: Matt D. R. <ya...@ap...> - 2002-11-15 07:15:02
|
On Thu, 14 Nov 2002, Randy.Dunlap wrote: |>Hi, |> |>I'm trying to merge lkcd into a 2.5.47 kernel. |>The config.in -> Kconfig changes are done; that's no problem. |> |>I am having one problem, and it's not really a merge problem. |> |>I am also merging high-res-timers (HRT). HRT works (passes |>its simple test programs) before I merge lkcd. |>After I merge lkcd & kprobes & LTT, the HRT test programs fail |>and the system clock is reset to Dec. 31, 1969. |> |>I'm trying to locate where this problem is coming from, and |>I want to build a kernel with CONFIG_CRASH_DUMP turned off. |>I'm not trying to say that lkcd is part of the problem -- |>I'm looking for a way to disable lkcd completely so that it |>is out of the picture. |> |>Well, I turn off CRASH_DUMP, but there's still some crashdump code |>in the kernel, and that's not a good thing. When trying to isolate |>problems, config options need to be able to eliminate code |>(and preferably data) completely, and I don't quite see LKCD |>doing that. Am I wrong about that? I'd like to be. |>Is someone already working on making this happen? If you want to completely eliminate the dump code, and assuming the module isn't loaded, just change the dump() macro to { }. I can't think of much else that would really affect you, with the exception of the PageInUse tags. We left things like dump_ipi(), dump_in_progress, etc., as standard parts of the kernel simply because we didn't anticipate that the patches would be rejected by Linus. That being said, we may have to go back to adding a couple of #if tags given where we're at. All that being said, I really doubt there is anything in the code with the exception of the page alloc patches that would be affecting HRT. It seems like LTT would the first candidate for analysis. If I can help you walk through what our patches contain to get the dump stuff out, let me know. --Matt P.S. I work on this in the evenings, so I can't always get to LKCD during the day given my day job. :) Sorry for the delay. |>(I can do this in my source tree for my testing purposes, but I |>really think it needs to be done in the lkcd sources.) |> |>Thanks, |> -- |