__color__ ticket summary component version milestone type owner status created _changetime _description _reporter 3 45 a way to bypass file read permissions using fuse client: Kernel module latest defect parl new 2012-03-14T10:41:37-0400 2012-03-14T10:41:37-0400 "This bug was reported by Andrew Savchenko, bircoph@gmail.com... Hello, On Mon, 12 Mar 2012 19:56:14 -0400 Kevin Stone wrote: > The invalid argument on chmod in fuse is fixed here: > > http://www.orangefs.org/trac/orangefs/changeset/9212 Thanks, it works. Though I found a way to bypass file read permissions using fuse client: 1) Mount pvfs2 using pvfs2-client. 2) do as root: # touch secret_file # chmod 0600 secret_file # echo secret_data > secret_file 3) do as user: $ pvfs2fuse -o fs_spec=""tcp://host:3334/pvfs2-fs"" mountpoint $ ll secret_file -rw------- 0 root root 12 Mar 13 19:58 secret_file $ cat secret_file secret_data Parent directory is owned by root. This works only for data read, you can't delete or modify file. pvfs2-client works normally. This doesn't work for directories, files only. With default_permissions fuse option this also works as expected. I know, security is not a primary goal for PFVS2 and most of distributed file systems, but nevertheless it will be good to fix this as well. Best regards, Andrew Savchenko " hubcap 3 2 acache thrashing Client-core 2.8.2 defect mtmoore new 2010-12-06T16:48:42-0500 2010-12-14T15:19:31-0500 "There is an issue with acache thrashing / effectiveness. ==== Main issues ==== 1. If optional attributes are part of a getattr request, and are not defined, when the client looks in the cache for the attribute it will consider the lack of optional attributes as a cache miss. 2. The size attribute is retrieved by itself and causes the other cached attributes to be invalidated (since any update to the acache invalidates the existing entry). ==== Possible changes ==== * no attribute masks used to confirm attributes in the cache * modify the acache update behavior to merge existing entries and new entries * remove the delineation between static and non-static attributes" mtmoore 4 25 Add Berkeley DB cache size to default configuration Admin apps latest enhancement parl new 2011-04-21T16:41:38-0400 2011-07-03T12:49:38-0400 Modify pvfs2-genconfig to include a reasonable value for DBCacheSizeBytes so servers by default use some amount of cache. mtmoore 4 9 Add SNMP MIB support Server 2.8.2 enhancement mtmoore new 2010-12-06T20:02:58-0500 2011-07-03T12:03:38-0400 Add support for SNMP queries against OrangeFS filesystems. mtmoore 5 5 change kernel logging to fit on a single line BMI 2.8.2 enhancement mtmoore new 2010-12-06T19:54:22-0500 2011-05-31T04:51:17-0400 Some kernel log messages include newlines which make grepping for log messages difficult. Change log messages to not include new lines. mtmoore 4 18 Client ignores ClientRetryDelayMilliSecs BMI latest defect mtmoore new 2011-02-03T18:55:19-0500 2011-02-03T18:55:19-0500 When the source is configured using --with-mx clients ignore ClientRetryDelayMilliSecs configuration option even though client's are using TCP (over myri interface). mtmoore 4 1 Handling of MTIME Requests/distributions 2.8.2 enhancement parl new 2010-11-12T10:58:23-0500 2010-11-12T10:58:23-0500 Handling of MTIME has always been an issue. Need to investigate the current behavior and propose a scalable alternative that gives some functionality compared to the POSIX spec walt 4 42 Investigate why Palmetto PVFS client got Filldir error Kernel module latest task parl new 2011-10-11T12:51:57-0400 2011-10-11T12:51:57-0400 "Sep 29 10:30:26 node0144 kernel: Filldir failed on one of the first two true PVFS directory entries. Sep 29 10:30:26 node0144 kernel: Duplicate entries may appear. Corey found the above messages in the logs. Don't know if this is a problem or not but should be investigated to find out." ligon 4 36 IOzone on FUSE mount fails when not root Admin apps latest defect parl new 2011-06-14T08:23:35-0400 2011-06-14T08:23:35-0400 "The iozone test done during nightly tests fails on Fuse mounts. The issue appears to be related to the file open with mode 0 and the subsequent truncate call. If iozone is modified to specify an 'appropriate' mode the test succeeds. Also, the test succeeds when ran as root. The permission check on the client side indicates it is denying access to truncate the file even though the user running the test is the file owner. Fuse mount with: bin/pvfs2fuse -o fs_spec=tcp://:3335/pvfs2-fs,allow_other,umask=022 /tmp/pvfs2-fuse-mount iozone: ./iozone -a -y 4096 -q $((1024*16)) -n 4096 -g $((1024*16*2)) -f /tmp/pvfs2-fuse-mount/iozone.out Current work around is to run the test as root for FUSE mounts" mtmoore 4 19 IPv6 support for BMI modules BMI latest task mtmoore new 2011-02-15T14:07:30-0500 2011-07-03T12:46:28-0400 Add IPv6 support to BMI modules that it is applicable to. mtmoore 3 37 kernel panic in pvfs2_devreq_writev Kernel module latest defect parl new 2011-06-24T08:36:16-0400 2011-06-24T08:36:16-0400 "Jun 18 11:36:44 user001 kernel: Unable to handle kernel paging request at 0000000000100100 RIP: Jun 18 11:36:44 user001 kernel: [] :pvfs2:pvfs2_devreq_writev+0x177/0x7b4 Jun 18 11:36:44 user001 kernel: PGD 603e379067 PUD 603f354067 PMD 0 Jun 18 11:36:44 user001 kernel: Oops: 0000 [1] SMP Jun 18 11:36:44 user001 kernel: last sysfs file: /devices/pci0000:00/0000:00:07.0/0000:0b:00.0/irq Jun 18 11:36:44 user001 kernel: CPU 0 Jun 18 11:36:44 user001 kernel: Modules linked in: pvfs2(U) dell_rbu ipv6 xfrm_nalgo crypto_api mx_driver(PU) mx_mcp(PU) ipmi_devintf ipmi_si ipmi_msghandler nfs fscache nfs_acl lockd sunrpc ipt_REJECT xt_state ip_conntrack nfnetlink xt_tcpudp iptable_filter ip_tables x_tables loop dm_mirror dm_multipath scsi_dh video backlight sbs power_meter hwmon i2c_ec i2c_core dell_wmi wmi button battery asus_acpi acpi_memhotplug ac parport_pc lp parport joydev sr_mod cdrom sg hpilo netxen_nic serio_raw pcspkr dm_raid45 dm_message dm_region_hash dm_log dm_mod dm_mem_cache mptsas(U) scsi_transport_sas mptspi(U) mptscsih(U) mptbase(U) scsi_transport_spi ata_piix libata shpchp cciss sd_mod scsi_mod ext3 jbd uhci_hcd ohci_hcd ehci_hcd Jun 18 11:36:44 user001 kernel: Pid: 10980, comm: pvfs2-client-co Tainted: P 2.6.18-194.17.1.el5 #1 Jun 18 11:36:44 user001 kernel: RIP: 0010:[] [] :pvfs2:pvfs2_devreq_writev+0x177/0x7b4 Jun 18 11:36:44 user001 kernel: RSP: 0018:ffff81603ea4ddd8 EFLAGS: 00010246 Jun 18 11:36:44 user001 kernel: RAX: 0000000000000000 RBX: 0000000000100100 RCX: 0000000000005143 Jun 18 11:36:44 user001 kernel: RDX: 000000000000011c RSI: ffff81156785e740 RDI: ffff81603ea4de20 Jun 18 11:36:44 user001 kernel: RBP: ffff812040102300 R08: 0000000000000000 R09: 0000000000000000 Jun 18 11:36:44 user001 kernel: R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000100100 Jun 18 11:36:44 user001 kernel: R13: 00000000000011c0 R14: ffff810ac03d03b8 R15: 0000000000000004 Jun 18 11:36:44 user001 kernel: FS: 00002ba44d671cc0(0000) GS:ffffffff803ca000(0000) knlGS:0000000000000000 Jun 18 11:36:44 user001 kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 Jun 18 11:36:44 user001 kernel: CR2: 0000000000100100 CR3: 000000603fa50000 CR4: 00000000000006e0 Jun 18 11:36:44 user001 kernel: Process pvfs2-client-co (pid: 10980, threadinfo ffff81603ea4c000, task ffff8140b07970c0) Jun 18 11:36:44 user001 kernel: Stack: ffff8140b07970c0 ffff81603ea4de88 000020306d38f76a ffff810ac03d03c0 Jun 18 11:36:44 user001 kernel: ffff8140b07972a8 0000000000000282 ffffffff8049bf80 0000000000000000 Jun 18 11:36:44 user001 kernel: 0000000000000000 000000000e22815e ffff81677c743a30 0000000000000000 Jun 18 11:36:44 user001 kernel: Call Trace: Jun 18 11:36:44 user001 kernel: [] do_readv_writev+0x172/0x291 Jun 18 11:36:44 user001 kernel: [] sys_writev+0x45/0x93 Jun 18 11:36:44 user001 kernel: [] system_call+0x7e/0x83 Jun 18 11:36:44 user001 kernel: Jun 18 11:36:44 user001 kernel: Jun 18 11:36:44 user001 kernel: Code: 4d 8b 24 24 48 8b 45 00 4c 01 e8 48 39 c3 75 c5 c7 45 20 01 Jun 18 11:36:44 user001 kernel: RIP [] :pvfs2:pvfs2_devreq_writev+0x177/0x7b4 Jun 18 11:36:44 user001 kernel: RSP Jun 18 11:36:44 user001 kernel: CR2: 0000000000100100 " mtmoore 3 44 merge BMI BMI latest task parl new 2011-10-24T08:46:42-0400 2011-10-24T08:46:42-0400 "BMI code has been updated and placed in a GIT repository by ANL at http://trac.mcs.anl.gov/projects/bmi. Merge in updates." mtmoore 4 10 mx performance is less than expected BMI 2.8.2 enhancement mtmoore new 2010-12-06T20:07:21-0500 2010-12-06T20:07:35-0500 Overall performance over MX is not as high as expected. mtmoore 4 17 openSUSE 11.2 invalid module format Kernel module 2.8.2 defect mtmoore assigned 2011-01-11T07:45:02-0500 2011-07-03T12:04:03-0400 "Per pvfs2-users list: http://www.beowulf-underground.org/pipermail/pvfs2-users/2011-January/003234.html pvfs2 2.8.2 on an openSuSE 11.2 with a 2.6.31.14-0.6 kernel creates a module that will not load, it throws: FATAL: Error inserting pvfs2 (/lib/modules/2.6.31.14-0.6-default/kernel/fs/pvfs2/pvfs2.ko): Invalid module format. " mtmoore 4 40 Path length should allow 4096 bytes instead of 256. Admin apps latest enhancement parl new 2011-08-04T16:13:30-0400 2011-08-04T16:13:30-0400 The client code currently allows only 256 bytes when specifying pathname + filename on the command line. We need a new PVFS_PATH_MAX_BYTES to identify the longest path length allowed. Currently, PVFS_NAME_MAX is set to 256, which is the correct value for a filename or a directory segment; however, this length was also used in some of the client code to specify total length of given path + filename. ligon 5 6 pvfs2-genconfig reports static path to pvfs2-server Server 2.8.2 enhancement mtmoore new 2010-12-06T19:57:32-0500 2010-12-06T19:57:32-0500 pvfs2-genconfig may be reporting a statically defined path to the pvfs2-server executable. Change it to report the real path of the executable. mtmoore 3 39 pvfs2-server fails to terminate Server latest defect parl new 2011-07-05T11:23:39-0400 2011-07-05T11:23:39-0400 "When given a kill signal the pvfs2-server process fails to terminate. Backtrace of the process is: 0x0000003e8c80d4c4 in __lll_lock_wait () from /lib64/libpthread.so.0 (gdb) bt #0 0x0000003e8c80d4c4 in __lll_lock_wait () from /lib64/libpthread.so.0 #1 0x0000003e8c808e1a in _L_lock_1034 () from /lib64/libpthread.so.0 #2 0x0000003e8c808cdc in pthread_mutex_lock () from /lib64/libpthread.so.0 #3 0x0000000000479039 in gen_posix_mutex_lock () #4 0x000000000043d261 in PINT_thread_mgr_bmi_stop () #5 0x00000000004346fc in job_finalize () #6 0x0000000000414147 in ?? () #7 0x00000000004122c8 in main () " mtmoore 5 30 pvfs2-touch and touch create files with different permissions.... Admin apps latest defect parl new 2011-05-05T16:54:16-0400 2011-07-03T12:50:22-0400 "pvfs2-touch creates a file with 777 permissions. touch creates a file with 644 permissions. Is this a problem for anyone? Should pvfs2-touch result in the same permissions as touch?" ligon 3 32 pvfs2-xattr segfault, cleanup Admin apps latest defect parl new 2011-05-18T03:55:10-0400 2011-05-18T03:55:10-0400 "pvfs2-xattr will segfault when doing a string comparison for the user.pvfs2.mirror* keys. Also, the needs badly to be re-worked for it to follow coding conventions and move semantic checks of key/value pairs user.pvfs2.mirror related attributes to the appropriate state machine." mtmoore 3 43 qlogic IB RTR->RTS invalid argument BMI latest defect parl new 2011-10-21T16:59:34-0400 2011-10-21T16:59:34-0400 "Using QLogic OFED (not 100% sure when it's an issue other than with QLogic gear) an error similar to the follow occurs: ""Error: init_connection_modify_qp: ibv_modify_qp RTR -> RTS: Invalid argument."" Patch to resolve the issue is attached. These parameters 1) should maybe be arguments 2) the MTU should be configurable? May want to check the BMI git repo for an update." mtmoore 4 35 Review LTP tests ran during nightly tests Admin apps latest enhancement parl new 2011-06-12T09:57:23-0400 2011-06-12T09:57:23-0400 The nightly tests include running a specific set of LTP tests. Review the tests available with LTP 20110606 and see if we should be running any others. mtmoore 4 16 setuid fails Admin apps latest defect mtmoore new 2010-12-19T09:42:47-0500 2011-07-03T12:11:45-0400 "Trying to chmod a file via the kernel inteface or pvfs2-chmod to 6XXX leads to an Invalid argument error. Setting only the gid bit (2XXX) succeeds. " mtmoore 5 8 test berkeley 11g Trove 2.8.2 task mtmoore new 2010-12-06T20:01:10-0500 2010-12-06T20:01:10-0500 Test OrangeFS against the new Berkeley DB (11g). mtmoore 3 33 The nightly tests fail when using the MX protocol Protocol/endecode latest defect parl new 2011-06-07T17:18:19-0400 2011-07-03T12:01:06-0400 "The dbench test loops endlessly while in the ""cleanup"" step. Not sure why this is happening and needs some investigation. I suspect that our mx module is just not behaving properly, in general." ligon 4 26 thread stack size Server latest enhancement mtmoore new 2011-04-27T06:39:24-0400 2011-04-27T06:39:24-0400 For short lived threads with few function calls we should shorten the stack to a known acceptable size instead of inheriting the value from ulimit. mtmoore 4 22 Unable to remove kernel module Kernel module latest defect mtmoore new 2011-03-24T08:21:01-0400 2011-07-03T11:55:15-0400 "Under certain conditions the kernel module is not able to be unloaded even though no mount exists and pvfs2-client is not running. The conditions that lead to this appear to be that: 1) the kernel module is loaded 2) pvfs2-client is running 3) a mount command is issued 4) the mount command doesn't complete (is hanging) and another insmod/pvfs2-client/mount sequence occurs. The end result is the module reference count is incorrect." mtmoore 4 28 Wikipedia documentation Server latest task parl new 2011-05-02T22:47:36-0400 2011-07-03T12:50:43-0400 Update wikipedia documentation with accurate information regarding OrangeFS. mtmoore