Hartmut Reuter wrote:
Well, I have now after Rainer's patch been able to write and read a 5 gb
file and it seems to work. But I am not sure that all kernel versions
support this. Therefor 5 years ago I chose this - may be ugly or
complicated - code to have large file support under linux which always
worked.
OK, I can confirm that once page_offset got defined correctly (but the
latest version of the patch using just the loff_t cast should solve this
just fine) I was able to copy an ~8G file (actually, a software library
volume dump with lots of entropy) into AFS and read it back via 'cksum'.
The good news: the cksum of the original and the file in AFS match!
Performance was lousy, though. Every now and then during the write the
client would hammer the fileserver for almost a half a minute with
zero-length FetchData RPCs (!), confirmed by fstrace on the client. I
vaguely remember having already seen this, though.
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Rainer Toebbicke
European Laboratory for Particle Physics(CERN) - Geneva, Switzerland
Phone: +41 22 767 8985 Fax: +41 22 767 7155
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel
