FreeBSD src-all

More graphs

2017-03-28 06:20:04.081502316086mav

MFC r313568 (by ken):Change the isp(4) driver to not adjust the tag type for REQUEST SENSE.The isp(4) driver was changing the tag type for REQUEST SENSEcommands to Head of Queue, when the CAM CCB flagCAM_TAG_ACTION_VALID was NOT set. CAM_TAG_ACTION_VALID is setwhen the tag action in the XPT_SCSI_IO is not CAM_TAG_ACTION_NONEand when the target has tagged queueing turned on.In most cases when CAM_TAG_ACTION_VALID is not set, it is becausethe target is not doing tagged queueing. In those cases, trying tosend a Head of Queue tag may cause problems. Instead, default tosending a simple tag.IBM tape drives claim to support tagged queueing in their standardInquiry data, but have the DQue bit set in the control mode page(mode page 10). CAM correctly detects that these drives do notsupport tagged queueing, and clears the CAM_TAG_ACTION_VALID flagon CCBs sent down to the drives.This caused the isp(4) driver to go down the path of setting thetag action to a default value, and for Request Sense commands only,set the tag action to Head of Queue.If an IBM tape drive does get a Head of Queue tag, it rejects it withInvalid Message Error (0x49,0x00). (The Qlogic firmware translates thatto a Transport Error, which the driver translates to an UnrecoverableHBA Error, or CAM_UNREC_HBA_ERROR.) So, by default, it wasn't possibleto get a good response from a REQUEST SENSE to an FC-attached IBMtape drive with the isp(4) driver.IBM tape drives (tested on an LTO-5 with G9N1 firmware and a TS1150with 4470 firmware) also have a bug in that sending a command with anon-simple tag attribute breaks the tape drive's Command ReferenceNumber (CRN) accounting and causes it to ignore all subsequentcommands because it and the initiator disagree about the nextexpected CRN. The drives do reject the initial command with a headof queue tag with an Invalid Message Error (0x49,0x00), but after thatthey ignore any subsequent commands. IBM confirmed that it is a bug,and sent me test firmware that fixes the bug. However tape drives inthe field will still exhibit the bug until they are upgraded.Request Sense is not often sent to targets because most errors arereported automatically through autosense in Fibre Channel and othermodern transports. ("Modern" meaning post SCSI-2.) So this is notan error that would crop up frequently. But Request Sense is useful ontape devices to report status information, aside from error reporting.This problem is less serious without FC-Tape features turned on,specifically precise delivery of commands (which enables CommandReference Numbers), enabled on the target and initiator. WithoutFC-Tape features turned on, the target would return an error andthings would continue on.And it also does not cause problems for targets that do taggedqueueing, because in those cases the isp(4) driver just uses thetag type that is specified in the CCB, assuming theCAM_TAG_ACTION_VALID flag is set, and defaults to sending a Simpletag action if it isn't an ordered or head of queue tag.sys/dev/isp/isp.c:In isp_start(), don't try to send Request Sense commandswith the Head of Queue tag attribute if the CCB doesn'thave a valid tag action. The tag action likely isn't validbecause the target doesn't support tagged queueing.
2017-03-16 11:04:13.940838315414pfg

Record mergeinfo for r310171.(From dim@:)Add __scanflike attributes to the kernel's sscanf() and vsscanf()declarations. This should help to catch future mismatches betweenformat strings and arguments.These were brought in r315282 by accident.
2017-03-11 03:11:08.638638315061dchagin

Record mergeinfo for r314217, r314218.
2017-03-08 22:22:01.136943314941kp

MFC r314810:pf: Fix a crash in low-memory situations(Merge-tracking only. This was mistakenly committed directly to stable/11 inr314702)
2017-02-22 15:11:41.499599314108dchagin

Record mergeinfo for r313284, r313285, r313684, r313912 missied in r314107.
2017-02-12 02:09:31.874896313678delphij

Fixup mergeinfo.
2017-02-10 01:58:40.552648313534delphij

MFC r274583: Default to use 10 seconds as nap interval instead of 1.(the mergeinfo portion; actual code changes were committed from wrongdirectory).
2017-02-09 20:59:53.271862313503ngie

Record mergeinfo for r288682 and r288683 (items that should never be MFCed)No net change
2017-02-09 16:37:55.871659313484ngie

Record r286620 and r286638 as having been MFCedNo net change
2017-02-03 07:52:02.492068313155des

More property fixups. Damn svn autoprops :(
2017-02-03 07:49:10.625724313153des

Add missing properties
2017-01-15 04:31:27.983055312217ngie

Delete svn mergeinfo for r312193 so I can once again sync^/projects/netbsd-tests-upstream-01-2017 with ^/head
2016-12-12 23:24:58.98195310001dteske

Merge missing mergeinfo for r303877
2016-12-04 10:59:22.192627309535jilles

Add missing mergeinfo for r309533.
2016-11-25 17:33:30.02358309169dim

Get rid of separate Subversion mergeinfo properties for llvm-dwarfdumpand llvm-lto. The mergeinfo confuses Subversion enormously, and thesedirectories will just use the mergeinfo for llvm itself.
2016-11-24 05:04:16.900534309094avg

MFV r308986: 6412 zfs receive: -u can be ignored sometimesThis is a record only change, the actual change is in FreeBSD sincer297520.
2016-11-02 19:43:48.33805308231jhibbits

Fix the build. protctl is only used on powerpc.While here, remove the need to check the SVR SPR, as others may be compatiblewith the p1022-esdhc type.
2016-10-20 14:28:31.106732307681rstone

Set the executable bit on arcgit and importgitMake it possible to run these scripts directly out of svn by settingthe executable property on them.
2016-09-22 12:05:43.743391306199imp

Revert svn:mergeinfo added inadvertantly in last commit r306197
2016-09-21 04:13:07.968803306079kib

Record mergeinfo for the r305971 commit, merge of r305380