﻿__group__	ticket	summary	component	version	type	owner	status	created	_changetime	_description	_reporter
	32	Superpages are not used as often as desired	kernel		defect	john	started	2013-06-07T15:20:45-04:00	2013-08-16T17:15:51-04:00	"At work we've found that there are some use cases where superpages are not used as often as we'd like.  Specifically, if we create several shared memory objects (via shm_open()) that are not exact multiples of a superpage and map them, they are assigned virtually contiguous addresses, meaning that they do not each start on a superpage boundary. This seems to be related to the fact that we only specify VMFS_ALIGNED_SPACE for device VM objects.

Secondly, if we pre-zero a file before mapping it, the file uses random pages from all over the place.  This may be harder to fix though alc@ may have a suggestion that might help."	john
	3	Handle resource leaks from driver detach routines	kernel		defect	john	assigned	2013-05-15T17:46:14-04:00	2013-06-27T17:13:43-04:00	Release active resources when a device is detached.	john
	5	Using %cr8/TPR for spinlock_enter/exit	kernel		defect	john	started	2013-05-15T18:04:00-04:00	2013-05-15T18:37:41-04:00	I have a branch of this implemented.  The trick so far is finding a workload where it helps.	john
	13	Fix concurrent RX processing in ixgbe(4)	kernel		defect	john	started	2013-05-15T22:35:38-04:00	2013-08-14T08:01:57-04:00	"There are at least two patches:
- Removing 'rearm_queues'
- Reworking watchdog handling (possibly needed in igb as well)"	john
	16	Better fix for UDP sends slow	kernel		defect	john	assigned	2013-05-15T22:56:10-04:00	2013-05-18T13:58:54-04:00	Figuring out the best way to close the race when you don't get the trylock when using drbr in an if_transmit routine.	john
	21	Make WITHOUT_GNU work	userland		defect	john	assigned	2013-05-15T23:00:42-04:00	2013-05-18T13:59:44-04:00	The WITHOUT_GNU src.conf option is currently a nop.  It should function as an installworld target that doesn't install any GNU binaries (e.g. src/gnu).	john
	30	hw.pci.realloc_bars is not quite right	kernel		defect	john	new	2013-05-30T11:12:44-04:00	2013-05-30T11:13:27-04:00	We should probably let a PCI bus allocate as many resources as possible using the firmware-assigned ranges and only fall back to reallocating BARs that failed in a second pass.	john
	33	Fix stdio to use int for fds instead of a short	userland		defect	john	started	2013-06-19T11:47:40-04:00	2015-10-22T17:50:36-04:00	Right now stdio uses a short for file descriptors instead of an int.  This means it can't handle FD's above 65535.	john
	55	Non-ISA PCI-PCI bridge behind a ISA PCI-PCI bridge can't allocate I/O window	kernel		defect		new	2014-03-26T12:07:41-04:00	2014-03-26T14:06:32-04:00	Will bad hardware never cease?	john
	64	Test rman_reserve_resource() with count of 0	boot		defect		new	2014-10-02T14:38:38-04:00	2014-10-02T14:39:09-04:00	Reserving a count of zero should fail.  Make sure it fails.	john
	97	Convert ticks to unsigned	kernel		defect		new	2015-03-05T07:13:46-05:00	2015-03-05T07:13:46-05:00		john
	1	multipass new-bus probe & attach	kernel		enhancement	john	started	2013-05-15T17:26:04-04:00	2014-01-21T12:14:59-05:00	Continuing the multipass probe & attach work to the point that we can start the full scheduler with timers before most devices are probed and attached.	john
	4	Reworking interrupt threads	kernel		enhancement	john	started	2013-05-15T17:56:46-04:00	2013-05-15T18:38:22-04:00	Tearing interrupt threads apart separating the queue of active handlers from the threads and allowing handlers to requeue themselves.  Also enables filters unconditionally.	john
	6	posix_fadvise() FADV_WILLNEED	kernel		enhancement	john	started	2013-05-15T18:09:45-04:00	2013-05-15T18:38:38-04:00	Adding some support for POSIX_FADV_WILLNEED to schedule read-ahead.  I have various patches to do this on UFS.  The naive version only triggered 16k reads.  I tried hacking up the clustering code to allow these to cluster but that was quite complex and still hard to get correct, so I think it is worth dropping.	john
	14	Try to reuse VM object lock in pagedaemon's page scan	kernel		enhancement	john	assigned	2013-05-15T22:46:28-04:00	2013-05-15T22:49:47-04:00	This was another pagedaemon change suggested by alc@.  One trick is figuring out a good way of developing a test case and a measurement to see if this makes any difference.	john
	19	Export CPU / NUMA topology to userland	kernel		enhancement	john	new	2013-05-15T22:58:34-04:00	2013-11-07T13:06:38-05:00	"One possible API suggested by brooks@:

http://www.open-mpi.org/projects/hwloc/doc/"	john
	20	Port repyter to FreeBSD	kernel		enhancement	john	assigned	2013-05-15T22:59:32-04:00	2013-05-18T13:59:13-04:00	"Source for repytr is here:

https://github.com/nelhage/reptyr"	john
	23	Add something like ipcs/ipcrm for POSIX IPC	userland		enhancement	john	assigned	2013-05-15T23:03:14-04:00	2013-05-18T14:00:29-04:00	At the very least it would be nice to list currently open shared memory objects and semaphores and provide a way to delete them.	john
	26	Port the broadcom wl driver	kernel		enhancement	john	accepted	2013-05-18T14:45:24-04:00	2013-05-20T08:28:16-04:00	Broadcom has a Linux driver with a binary blob and source (for Linux) for several broadcom wireless parts.	john
	40	Tools to evaluate NUMA-related performance	userland		enhancement		new	2013-11-07T13:05:50-05:00	2013-11-07T13:05:50-05:00	Linux has a tool called numatop and a daemon called numad that we should investigate and possibly mimic.	john
	96	SMBIOS 64-bit table support	kernel		enhancement		new	2015-03-05T06:55:26-05:00	2015-03-05T06:55:26-05:00	"Also need to export an SMBIOS table interface for this that the IPMI driver can use instead of having duplicate table parsing code.

Would need to update the table parsing code in the loader as well as in the kernel."	john
	15	Rename boot2 to ufsboot	boot		task	john	assigned	2013-05-15T22:54:14-04:00	2013-05-18T13:58:35-04:00	Rename sys/i386/boot2 to sys/i386/ufsboot and /boot/boot to /boot/ufsboot.	john
	38	Write getenv(9) and TUNABLE(9)	manpages		task	john	started	2013-10-04T15:30:56-04:00	2013-10-22T16:31:21-04:00	Manpages for the in-kernel environment variable API as well as the TUNABLE API which sits on top of it.	john
	39	psutil.py issues	userland		task		new	2013-10-15T10:44:02-04:00	2013-10-15T10:44:02-04:00	"http://code.google.com/p/psutil/issues/detail?id=373

May also look for other issues in psutil that are FreeBSD related while I'm there?"	john
	57	Investigate using SSE/AVX for amd64 libc routines	userland		task	john	accepted	2014-05-06T14:22:45-04:00	2014-05-06T14:23:27-04:00	Right now libc contains assembly versions of several commonly used routines like memcpy(), etc.  These routines generally do not attempt to use SSE but rely on older primitives like 'rep movs'.	john
	61	Retire swi entirely	kernel		task	john	new	2014-09-19T11:10:45-04:00	2014-09-19T11:10:45-04:00	Just use plain kthreads or taskqueues for remaining swi handlers.  They can use wakeup in place of swi_sched().  Having swi's adds extra complication to the interrupt code and don't really serve a distinct purpose without spls.	john
	62	Import extended timezone API	userland		task	john	new	2014-09-29T16:32:00-04:00	2016-01-28T11:25:20-05:00	This was a local patch at HRT, but the tzcode folks have adopted an upstream version from NetBSD that does the same thing.  The task now would be to import the newer tzcode into stock FreeBSD and update manpages to document the new functions.	john
	63	Test TDF_SBDRY check	kernel		task	john	new	2014-10-02T14:29:20-04:00	2014-10-02T14:29:20-04:00	Add a check for interruptible sleeps holding locks that do not have TDF_SBDRY set and see if anything triggers that.  If it does, see if those are bugs to be fixed, or if this assertion is infeasible.	john
	90	Kill basemem	kernel		task		new	2015-01-23T18:21:31-05:00	2015-01-23T18:21:31-05:00	Remove the basemem variable from the x86 kernels and instead use loops over physavail/dumpavail.	john
	91	Re-port mcelog	userland		task		new	2015-01-23T18:22:37-05:00	2015-03-05T07:16:16-05:00	Re-port changes to add FreeBSD support to a git clone of the official mcelog sources.	john
	92	Add per-mount client NFS stats like mountstats	kernel		task		new	2015-01-23T18:23:48-05:00	2015-01-23T18:23:48-05:00	Similar to mountstats from Linux, track per-mount, per-RPC statistics including RTT, average bytes / operation, etc.	john
	95	Add a device_panic callback	kernel		task	john	new	2015-02-03T14:58:08-05:00	2015-02-03T14:58:08-05:00	The idea would be that drivers would get a chance to save off state during a panic that will end up in the crashdump.  Drivers would be responsible for pre-allocating stuff ahead of time, etc.	john
	98	Update machine check code to handle recoverable UC	kernel		task		new	2015-03-05T07:14:32-05:00	2015-03-05T07:14:32-05:00	In general compare the current machine check parsing against the latest SDM and add support for missing features, etc.	john
	7	Require busdma boundary parameters to be powers of 2	kernel		defect	john	assigned	2013-05-15T18:10:35-04:00	2013-05-15T22:50:06-04:00	Just fail non power of 2 values with EINVAL or some such.	john
	34	Fix pci_cfgreg*() to handle domains	kernel		defect		new	2013-07-24T12:36:20-04:00	2013-07-24T12:36:20-04:00	Add an explicit domain parameter to pci_cfgreg* and undo the hack for ia64 from r253560	john
	24	Retire swi_vm	kernel		enhancement	john	assigned	2013-05-16T10:14:28-04:00	2013-05-18T14:03:28-04:00	Replace it with an MD dma swi handler on platforms that use bounce pages.	john
