Lustre devel - Feb 2007 - [Bug 11290] ldiskfs changes to support server I/O cleanup

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11290



Created an attachment (id=9498)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
 --> (https://bugzilla.lustre.org/attachment.cgi?id=9498&action=view)
support for non-aligned appends via DIO

tested a bit ...

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11290



Created an attachment (id=9499)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
 --> (https://bugzilla.lustre.org/attachment.cgi?id=9499&action=view)
a silly programm to test non-aligned appends

issues few DIO appends, then checks consistency with buffered reads.

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11290



Created an attachment (id=9500)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
 --> (https://bugzilla.lustre.org/attachment.cgi?id=9500&action=view)
range locking support for ext3 (alpha version)

tested with concurrent appends on UP. need much more testing ... in this patch
generic range locks support is built into ext3, but the feature should go into
kernel''s lib/.

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11290



Created an attachment (id=9501)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
 --> (https://bugzilla.lustre.org/attachment.cgi?id=9501&action=view)
a program doing concurrent appends to test range locking for ext3

it especially exams a case when file is opened with O_APPEND and i_size changes

after we got lock.

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11290



Created an attachment (id=9508)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
 --> (https://bugzilla.lustre.org/attachment.cgi?id=9508&action=view)
fine grained cache invalidation for dio patch

the idea is that we don''t need to flush/invalidate a whole mapping.
rather we
should update affected range.