19:06:10 #startmeeting 19:06:10 Meeting started Sat Dec 5 19:06:10 2009 UTC. The chair is maxamillion. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:06:10 Useful Commands: #action #agreed #halp #info #idea #link #topic. 19:06:20 #Topic - Provice of the Kernel 19:07:18 Community of upstream Fedora and RedHat developers, who they are and what they do 19:07:35 new: 19:08:13 Perf Counters -> hardware performance counters, events, software events, perf top, inspecting userspace code to support reporting on "how your code is running" 19:08:39 Allows us to inspect the code at a microscopic and macroscopic level (explanation to come) 19:09:22 Can also "look at" userspace code and shared libraries 19:09:30 "Shiny Shiny": 19:09:36 -> usb3 support 19:09:53 -> wireless personal networks (not sure what it does, but the kernel supports it now) 19:10:37 -> kmemcheck and kmemleak, in the debug kernels it reports on different memory issues and will log as well as notify mailing lists upon events and such 19:10:53 -> fsnotify, will notify upon filesystem events 19:11:34 -> intel iommu support (out of tree) .... good for virtualization, creating security break up of memory spaces 19:11:42 "Holy moley KMS": 19:11:58 -> intel check .... went by too fast, sorry all 19:12:08 "Whats new in 2.6.32": 19:12:19 Currently in rawhide, will fork and be put into Fedora 12 19:12:32 will also be broken out into Fedora 11 19:12:45 Fedora 12 should be soon, Fedora 11 should be sometime next week 19:12:57 "Generic changes in 2.6.32": 19:13:39 -> KSM (kernel same page merging) ---> essentially data deduplication for RAM 19:15:45 -> hwpoison (fault injection for testing) 19:16:19 -> trim and discard (for modern SSD, increases filesystem performance) 19:16:36 -> tracing (perf counters are now perf events) 19:16:45 --> perf timechart (bootchart) 19:16:57 --> perf sched (scheduler latency tracing) 19:17:17 --> tracepoints (syscalls, modules, skb, pagealloc, etc...) 19:17:45 -> cgroups: soft memory limits 19:18:01 -> dynamic per cpu scalability 19:18:06 "Virt": 19:18:23 -> ioeventfd: performance enhancement for virtual machines for I/O 19:18:33 -> irqfd: interrupts for hosts 19:18:47 -> syscall/sysenter emulation 19:19:01 -> gigabyte pages: permance enhancing for ram 19:19:12 -> ... missed this one (sorry all) 19:19:18 "filesystems": 19:19:27 -> btrfs improvements: 19:19:40 --> -ENOSPC (Can handle running out of space now) 19:19:45 --> snapshots 19:19:49 --> deletion 19:19:54 --> performance 19:19:59 -> ext4 19:20:35 --> writeback improvements 19:20:42 --> async commit with barrier 19:20:46 --> fallocate 19:20:50 -> xfs 19:20:56 --> performance enhancements 19:20:59 "drivers": 19:21:18 -> b43 lpphy (dell netbooks and more for the low power broadcom) 19:21:29 -> ti wl12xx wireless 19:21:36 -> broadcom ssb sdio support 19:21:45 -> r600/r700 modesetting 19:22:05 -> moorestown/simple firmware interface (sfi) ---> atom based 19:22:30 -> mac80211 improvement (background scanning, more wet in cfg80211) 19:22:35 -> vmware vmxnet 19:22:49 -> ath9k (new chipsets), ath5k (hw crypto) 19:22:59 -> video4linux2/dvb improvements 19:23:04 -> acpi4 19:24:48 "staging" 19:25:04 -> rt3090/rt3070/rt2870 19:25:13 -> htc dream (android phone) 19:25:20 -> hyperv (microsoft virt) 19:26:04 "staging" refers to drivers that are "questionable" such that they aren't in the kernel upstream project, and as an effort from Greg Kroah-Hartman the staging tree was created for them 19:26:34 they aren't currently included in Fedora unless there is a maintainer willing to handle wrangling the code 19:26:59 these drivers are not included in the upstream kernel because their code is "questionable" (just a note for clarity) 19:27:45 "crystal ball" (upcoming 2.6.24 hopes and dreams): 19:28:05 -> poulsbo (intel gma500) support 19:28:10 -> more staging migration 19:28:21 -> rt2xxx/rt3xxx wifi 19:28:52 "how to get help?" 19:29:49 -> the fedora mailing lists and bugzilla are popular but the best way is on #fedora-devel or #fedora-kernel 19:30:47 this is just due to raw volume of bugs and other aspects of rawhide that make developers inboxes very bloated 19:32:11 "please test rawhide" -> rawhide is often difficult to run, wrangle, and survive but it is very helpful for kernel developers to have more hands in the testing world, often times you can just take the rawhide kernel build from koji and install it on a "fedora stable" system 19:32:28 #topic State of X 19:32:47 radeon: 19:32:57 -> kms for r600+ 19:33:05 -> r600+ 3d 19:33:14 -> r800 support coming soon 19:33:17 -> displayport 19:33:22 -> hdmi audio 19:33:47 -> xvmc for r300 through r500, 600+ soon 19:33:52 -> future 19:34:02 --> sideport scanout 19:34:09 --> page flipping 19:35:02 A lot of this (mainly output setup and memory management) is related to getting a block of memory in the kernel, and these are now being integrated (KMS - kernel mode setting) 19:36:58 the r800 chips were just recently released and this support is being worked on and the upstream kernel already has the commit of the code to implement that 19:40:44 #topic State of X - intel 19:40:49 -> displayport 19:41:22 -> pch changes (periferal(sp?) control hub) 19:41:38 -> future 19:41:45 --> gpu scheduling 19:41:56 --> zone rendering 19:42:04 --> per process context and gtt 19:42:10 -> poulsbo 19:42:14 -> larrabee 19:44:35 Working on having a GPU scheduler such that its not "competing" with the kernel scheduler (such that the kernel wants to get everything done as fast as possible all the time, but displays often have situations like "I want to guarantee 60 frames per second at all times 19:44:39 ) 19:46:41 #topic State of X - nvidia 19:46:59 -> nouveau with KMS (kernel mode setting) 19:47:07 -> 3d support (kinda) 19:48:36 * nirik notes 3d support will be very welcome for lots of people who currently resort to the nonfree driver. 19:48:42 #topic State of X - Core DRM (Digital Rights Management) 19:48:51 -> ddc handwaving 19:49:08 -> better multi-master (X server is an example of a master) 19:49:54 -> usb device support? --> couple of usb video devices (one end of cable is usb and the other is a vga or dvi connector) 19:51:27 usb devices is a difficulty of deciding design of how to make the native graphics card communicate with the usb device 19:51:33 #topic State of X - Xserver 19:51:40 -> xi2 19:51:47 -> randr and xinerama changes 19:51:55 -> multiseat 19:52:01 -> presentation and sync apis 19:52:06 -> devel model changes 19:52:14 -> even more embedded focus 19:57:50 #topic State of X - Mesa and Misc. 19:58:03 -> OpenGL3 support is being worked on 19:59:02 -> gallium? different build process that some devices have both gallium and classic drivers (which one do we pick?) 19:59:36 -> KMS plan for non-big-3 chips (AMD/nVidia/Intel) ... "What do we do with them?" 19:59:44 -> panic messages in X 19:59:50 -> headless polish 20:01:35 #endmeeting