Messages in this thread | | | Date | Sat, 25 Dec 1999 16:29:49 -0800 | From | "B. D. Elliott" <> | Subject | Re: [PATCH] 2.3.34/pre-2.3.35-3 ramdisk/initrd NOT as a |
| |
On Sat, Dec 25, 1999 at 03:28:37AM -0800, Tigran Aivazian wrote: > Dear Linus and the World, > > Merry Christmas! > > You said recently about BDE's ramdisk fix (included in Frank's patch): > > > Looks good, applied. > > I agree that it looks very good but it has one problem - it can easily > cause "VM: oom" events - just insmod rd with blocksize=512 and start > filling up the minors dd'ing a large file to the corresponding /dev/ramX > nodes. > > A simpler approach (using vmalloc()) as I suggested earlier does not > result in the disaster described above. Therefore, perhaps it still makes > sense to consider the patch below: > > http://www.ocston.org/~tigran/patches/ramdisk-2.3.35-pre2.patch > > Alternatively, if you really don't like vmalloc() idea one ought to > investigate how to avoid those ooms.
I haven't been able to reproduce this condition with the version that was submitted. On a 64MB machine I created 12 4MB ramdisks, did "mke2fs /dev/rd/$i", and filled all of them with "dd if=/dev/zero of=/rd.$i/big bs=1024 count=3933". There was some swapping while writing the final one, to make enough space. I also tried filling all of them by writing directly to each block device. Again there was some swapping while filling the last one.
The first version of the patch (which I think is still on your web site), had a bug which protected the wrong buffer in many/most cases. Perhaps this is what you are using?
If you can reproduce this with the current version, please let me know exactly how to do it.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |