1 ## Note: this is my personal todo and ideas list (alexh@)
3 * Add a simple (regression) test framework
4 - Add the hundreds of test cases from our bugtracker
5 - Run nightly/weekly or at least before a release
8 - add a callout_init_mtx/lock, etc
11 * Improve installer crypto support
12 - add tcplay/TrueCrypt support
13 - let user choose crypto algorithm, IV generation, etc
15 * Add support for hyperthreading / other SMT to our scheduler
16 - distinguish between real cores and threads on the same core
17 - Linux' work in the area: http://lwn.net/Articles/8553/
19 * Take advantage of nested paging/EPT in vkernels
21 * add a non-persistent unionfs VFS
22 - while this does not fully replace unionfs, it would deal with a few of the situations where unionfs is useful
23 - additionally the complexity is much lower, since everything can be kept in memory
25 * bring the samba3 hammer shadow copy foo to maturity
27 * extend dm's message feature to be two-way? what's the benefit?
29 * make the whole system deal correctly with non-512-byte blocksize for disks. (DEV_BSIZE)
31 * add another optimized bcopy/memcpy/memset/bzero version, using the AVX instructions
33 * cryptdisks: multiple keyfile support
35 * Add geli (without integrity verification) support using dm_target_crypt?
36 - a bit like the recent TrueCrypt implementation (tcplay)
40 * Keep opencrypto up-to-date
41 - http://www.openbsd.org/cgi-bin/cvsweb/src/sys/crypto/
43 * Port hwpmc & dig into (boot-up) performance
47 - separate out common arch parts (linprocfs, for example)
49 * Fix the crash analysis script (or rather the programs it calls [some segfault])
51 * Take a look at updating lvm/dm/libdevicemapper
53 * Take a look at importing dmctl from NetBSD
54 - would require a bunch of modifications to work with the proper device-mapper API
55 - also some extensions? missing lots of features
57 * rip out the disk partitioning from the disk subsystem and implement it in a more general fashion
58 - crazy idea: as dm targets with an auto-configuration option!
59 - would require to be able to create dm targets with an arbitrary name and not in /dev/mapper
61 * ATA (automatic) spindown (see FreeBSD current)
64 http://svn.freebsd.org/viewvc/base?view=revision&revision=127969
68 - some incorrect accounting going on, don't remember details :)
71 - make it work without whiteout
77 * Fix ipsec, get rid of old ipsec
80 - Added VT6105M specific register definitions. VT6105M has the following hardware capabilities.
81 - Tx/Rx IP/TCP/UDP checksum offload.
82 - VLAN hardware tag insertion/extraction. Due to lack of information
83 for getting extracted VLAN tag in Rx path, VLAN hardware support
84 was not implemented yet.
85 - CAM(Content Addressable Memory) based 32 entry perfect multicast/
88 o Implemented CAM based 32 entry perfect multicast filtering for
89 VT6105M. If number of multicast entry is greater than 32, vr(4)
90 uses traditional hash based filtering.
92 * RedZone, a buffer corruption protection for the kernel malloc(9) facility has been implemented.
93 - This detects both buffer underflows and overflows at runtime on free(9) and realloc(9),
94 and prints backtraces from where memory was allocated and from where it was freed.
97 * port uart driver (?)
99 * port wscons (?) or update syscons
100 - probably way too much effort (wscons)
104 - wrapper is included for userland; should be easy to port
105 - http://svn.freebsd.org/viewvc/base?view=revision&revision=184610
106 - http://turbocat.net/~hselasky/usb4bsd/
107 - http://gitweb.dragonflybsd.org/~polachok/dragonfly.git/shortlog/refs/heads/usb2
109 * suspend/resume for SMP x86
110 - http://lists.freebsd.org/pipermail/freebsd-acpi/2008-May/004879.html
112 * AMD64 suspend/resume
113 - http://svn.freebsd.org/viewvc/base?view=revision&revision=189903
120 [alexh@leaf:~/home] $ roundup-server -p 8080 bt=bugtracker
123 -05:48- : dillon@: no, double frees to the object cache are nasty. It can't detect them. the object
124 winds up in the magazine array twice
125 -05:48- : dillon@: (and possibly different magazines, too)
126 -05:49- : alexh@: can't I just write some magic to a free object on the first objcache_put and check
127 if it's there on objcache_put?
128 -05:49- : alexh@: and clear it on objcache_get, anyways
129 -05:50- : dillon@: no, because the object is still may have live-initialized fields
130 -05:50- : dillon@: because it hasn't been dtor'ed yet (one of the features of the objcache, to avoid
131 having to reinitialize objects every time)
132 -05:50- : dillon@: the mbuf code uses that feature I think, probably other bits too
133 -05:51- : dillon@: theoretically we could allocate slightly larger objects and store a magic number at
134 offset [-1] or something like that, but it gets a little iffy doing that
135 -05:52- : dillon@: the objcache with the objcache malloc default could probably do something like that
137 -05:52- : dillon@: I don't consider memory tracking to be a huge issue w/ dragonfly, though I like the
138 idea of being able to do it. It is a much bigger problem in FreeBSD due to the
139 large number of committers
142 -05:55- : dillon@: For the slab allocator you may be able to do something using the Zone header.
143 -05:55- : dillon@: the slab allocator in fact I think already has optional code to allocate a tracking
144 bitmap to detect double-frees
145 -05:56- : dillon@: sorry, I just remembered the bit about the power-of-2 allocations
146 -05:56- : dillon@: for example, power-of-2-sized allocations are guaranteed not only to be aligned on
147 that particular size boundary, but also to not cross a PAGE_BOUNDARY (unless the
149 -05:57- : dillon@: various subsystems such as AHCI depend on that behavior to allocate system
150 structures for which the chipsets only allow one DMA descriptor.
151 -05:59- : alexh@: http://svn.freebsd.org/viewvc/base/head/sys/vm/redzone.c?view=markup&pathrev=155086
152 < this is redzone. it basically calls redzone_addr_ntor() to increase the size in
153 malloc(), and then redzone_setup() just before returning the chunk
154 -06:02- : dillon@: jeeze. that looks horrible.
155 -06:03- : alexh@: I don't quite get that nsize + redzone_roundup(nsize)
156 -06:03- : dillon@: I don't get it either. It would completely break power-of-2-sized alignments in the
158 -06:04- : dillon@: hmmm. well, no it won't break them, but the results are oging to be weird
159 -06:04- : dillon@: ick.
161 -06:15- : dillon@: if the original request is a power of 2 the redzone adjusted request must be a power
163 -06:15- : dillon@: basically
164 -06:16- : dillon@: so original request 64, redzone request must be 128, 256, 512, 1024, etc.
165 -06:16- : alexh@: yah, k
166 -06:16- : dillon@: original request 32, current redzone code would be 32+128 which is WRONG.
167 -06:16- : alexh@: how big is PAGE_SIZE ?
168 -06:16- : dillon@: 4096 on i386 and amd64
169 -06:17- : alexh@: and one single malloc can't be bigger than that?
170 -06:17- : dillon@: I'm fairly sure our kmalloc does not guarantee alignment past PAGE_SIZE (that is,
171 the alignment will be only PAGE_SIZE eve if you allocate PAGE_SIZE*2)
172 -06:17- : dillon@: a single kmalloc can be larger then PAGE_SIZe
173 -06:18- : dillon@: it will use the zone up to around 1/2 the zone size (~64KB I think), after which it
174 allocates pages directly with the kernel kvm allocator
175 -06:18- : dillon@: if you look at the kmalloc code you will see the check for oversized allocations
176 -06:18- : alexh@: yah, saw that
177 -06:18- : alexh@: "handle large allocations directly"
178 -06:19- : alexh@: not sure how to do this, really, as the size is obviously also changed in
180 -06:20- : alexh@: but kmem_slab_alloc isn't called always, is it?
181 -06:20- : alexh@: only if the req doesn't fit into an existant zone
182 -06:20- : dillon@: right
183 -06:20- : dillon@: you don't want to redzone the zone allocation itself