From de34f5ed5e0c3c20dc99ae9cefa0b201d49c95b5 Mon Sep 17 00:00:00 2001 From: Samuel Thibault Date: Fri, 11 Feb 2011 17:48:07 +0100 Subject: Add rm -fr slowness issue --- open_issues/rm_fr.mdwn | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+) create mode 100644 open_issues/rm_fr.mdwn diff --git a/open_issues/rm_fr.mdwn b/open_issues/rm_fr.mdwn new file mode 100644 index 00000000..89a803ab --- /dev/null +++ b/open_issues/rm_fr.mdwn @@ -0,0 +1,27 @@ +[[!meta copyright="Copyright © 2011 Free Software Foundation, Inc."]] + +[[!meta license="""[[!toggle id="license" text="GFDL 1.2+"]][[!toggleable +id="license" text="Permission is granted to copy, distribute and/or modify this +document under the terms of the GNU Free Documentation License, Version 1.2 or +any later version published by the Free Software Foundation; with no Invariant +Sections, no Front-Cover Texts, and no Back-Cover Texts. A copy of the license +is included in the section entitled [[GNU Free Documentation +License|/fdl]]."]]"""]] + +[[!tag open_issue_hurd]] + +From: Samuel Thibault +Subject: rm -fr slowness + +I have always been surprised by the slowness of a mere rm -fr. Looking a +bit inside, I see that diskfs_dirremove_hard() calls diskfs_file_update +(dp, 1) (as does diskfs_truncate, diskfs_direnter_hard, and +diskfs_dirrewrite_hard). diskfs_file_update then calls pager_sync on +the pager, which thus writes back the whole ext2fs pager! + +This sounds a bit excessive to me, an unlink could just record it in +memory and actually sync later. Also, the wait flag is set, so we +really waits for all I/Os, which basically means strictly serializing +file removals: remove one file, wait for the disk to have done it +(~10ms), remove the next one, etc. I guess this is for safety reasons +against crashes, but isn't the sync option there for such kind of -- cgit v1.2.3