src-test2/sys/dev/md, branch release/6.3.0_cvs FreeBSD source tree https://cgit-dev.freebsd.org/src-test2/atom?h=release%2F6.3.0_cvs 2008-01-15T15:46:22Z This commit was manufactured by cvs2svn to create tag 2008-01-15T15:46:22Z cvs2svn cvs2svn@FreeBSD.org 2008-01-15T15:46:22Z urn:sha1:5cd221028ba5e7ada35a0dd196fb642b44f197d1 'RELENG_6_3_0_RELEASE'. This commit was manufactured to restore the state of the 6.3-RELEASE image. MFC 2007-01-05T09:51:14Z Konstantin Belousov kib@FreeBSD.org 2007-01-05T09:51:14Z urn:sha1:4828c37e1bcb358a0e0648072314c9175f66da14 rev. 1.167 of src/sys/dev/md/md.c rev. 1.514, 1.515 of src/sys/kern/vfs_bio.c rev. 1.319, 1.320 of src/sys/sys/vnode.h Resolve two deadlocks that could be caused by busy md device backed by vnode. MFC of rev. 1.164: 2007-01-04T17:39:17Z Konstantin Belousov kib@FreeBSD.org 2007-01-04T17:39:17Z urn:sha1:b15ce2d3a65287061e71fbaa1f0663c2f346b28f - Conditionally acquire Giant in mdstart_vnode(), mdcreate_vnode(), and mddestroy() only if the file is from a non-MPSAFE fs. There, to accommodate for the vn_open() difference with HEAD, use VFS_LOCK_GIANT after vn_open (since vn_open(..., -1) assumes caller is not MP-safe). - No longer unconditionally hold Giant in the md kthread for vnode-backed kthreads. - Improve the handling of the thread exit race when destroying an md device. No objections from: jhb MFC: make sure the start and end markers are contiguous and 2006-02-14T14:46:22Z Luigi Rizzo luigi@FreeBSD.org 2006-02-14T14:46:22Z urn:sha1:2ec6443b688a7277f301d32f47bc615b2b6f0095 not suppressed by the optimizer. Approved by: re MFC: sys/dev/md/md.c 1.161 2006-01-30T23:03:44Z Pawel Jakub Dawidek pjd@FreeBSD.org 2006-01-30T23:03:44Z urn:sha1:22f4532b3dcae459bb5dae1d2d9052491872993e Call NDFREE() only when vn_open() succeeded. MFC: 2005-10-07T15:08:41Z Poul-Henning Kamp phk@FreeBSD.org 2005-10-07T15:08:41Z urn:sha1:31d14eb0d633662f149c55f264a22a9c67b53bbd make sure the type is set early enough for the kthread to know to grab Giant for vnode backed devices. Approved by: re@ (scottl) MFC: 2005-09-26T18:14:38Z Poul-Henning Kamp phk@FreeBSD.org 2005-09-26T18:14:38Z urn:sha1:5eec9ecc16af67af8b7df133c24c45c0fd837f6d Fixes for various races in md. Approved by: re@ (scottl) MFC md.c rev 1.154 2005-09-09T21:39:47Z Christian S.J. Peron csjp@FreeBSD.org 2005-09-09T21:39:47Z urn:sha1:47c3af75ee00b10a8d76d2ea52ad516682c87466 MFC mdconfig.c rev 1.44 Dont allow read only backing stores to be modified PR: kern/84635 Approved by: re (kensmith) MFC: sys/dev/md/md.c 1.155 2005-09-03T16:02:37Z Pawel Jakub Dawidek pjd@FreeBSD.org 2005-09-03T16:02:37Z urn:sha1:b70af692ed0624f2784af13b73c05db2ee42ba34 - Add md_mtx lock to protect ID number and list of devices. - Always check mdnew() return value, as even in !autounit case kthread_create() can fail. Those two changes fix serval panics provoked by simple stress test. Tested by: Kris The BugMagnet Approved by: re (scottl) Request a CPU private mapping from sf_buf_alloc(). If the swap-backed 2005-02-13T21:51:50Z Alan Cox alc@FreeBSD.org 2005-02-13T21:51:50Z urn:sha1:e340fc602be23c05a35f023bb91e70b4148c2ad9 memory disk is larger than the number of available sf_bufs, this improves performance on SMPs by eliminating interprocessor TLB shootdowns. For example, with 6656 sf_bufs, the default on my test machine, and a 256MB swap-backed memory disk, I see the command "dd if=/dev/md0 of=/dev/null bs=64k" achieve ~489MB/sec with the default, shared mappings, and ~587MB/sec with CPU private mappings.