huangwei@clusterfs.com
2007-Feb-06 03:12 UTC
[Lustre-devel] [Bug 11300] Interval trees for scalable extent lock handling
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by using the following link: https://bugzilla.lustre.org/show_bug.cgi?id=11300 Created an attachment (id=9523) Please don''t reply to lustre-devel. Instead, comment in Bugzilla by using the following link: --> (https://bugzilla.lustre.org/attachment.cgi?id=9523&action=view) DLD revision2 fixed defects from vitaly''s inspection. in addition: 1. for defect 4:>have a look at @conflicing variable and its logic. >it is completely skipped in your dld: you have to get @conflicting >from ldlm_extent_lock_policy or explain how this logic will be changed.seems "conflicting" variable is mostly used for waiting list; for original granted list case, because new_ex.start = 0 and new_ex.end = EOF, so "conflicting" variable need to be taken care of; but for current case, we can reassign new_ex to req->l_req_extent, then we no longer need to care of "conflicting" now 2. policy list maintainment in bug10902, @l_sl_policy is implemented as a not cycle list, so for extent interval tree case, we accept this and will not use any list_*() anymore.