Messages in this thread | | | Date | Wed, 06 Sep 2006 16:38:14 +1000 | From | Nick Piggin <> | Subject | Re: [Feature] x86_64 page tracking for Stratus servers |
| |
Kimball Murray wrote:
>Attached is a git patch that implements a feature that is used by Stratus >fault-tolerant servers running on Intel x86_64 platforms. It provides the >kernel mechanism that allows a loadable module to be able to keep track of >recently dirtied pages for the purpose of copying live, currently active >memory, to a spare memory module. > >In Stratus servers, this spare memory module (and CPUs) will be brought into >lockstep with the original memory (and CPUs) in order to achieve fault >tolerance. > >In order to make this feature work, it is necessary to track pages which have >been recently dirtied. A simplified view of the algorithm used in the kernel >module is this: > >1. Turn on the tracking functionality. >2. Copy all memory from active to spare memory module. >3. Until done: > a) Identify all pages that were dirtied in the active memory since > the last copy operation. > b) Copy all pages identified in 3a to the spare memory module. > c) If number of pages still dirty is less than some threshhold, > i. "black out" the system (enter SMI) > ii. copy remaining pages in blackout context > iii. goto step 4 > Else > goto 3a. >4. synchronize cpus >5. leave SMI, return to OS >6. System is now "Duplexed", and fault tolerant. >
Silly question, why can't you do all this from stop_machine_run context (or your SMI) that doesn't have to worry about other CPUs dirtying memory?
>Please consider this feature for inclusion in the kernel tree, as it is very >important to Stratus. >
Given that it doesn't touch core mm/ code, I don't really care about it[*] except that it doesn't make sense to have the tracking hooks in generic code because it is pretty specific to your module.
Also, as far as a "notifier" goes, it is of course completely broken unless your module is the only one that is ever going to use it. Which I suspect may be the case ;) But you do really need to at least WARN_ON or fail if someone is going to break it like this.
[*] Though if it gets included, it would not stop me lamenting the proliferation of complexities to support *tiny* obscure userbases. Can we wait until your hardware is smart enough to snoop the cc? :)
--
Send instant messages to your online friends http://au.messenger.yahoo.com
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |