FreeBSD cvs-src summary for 01/11/04 to 08/11/04 ++++++++++++++++++++++++++++++++++++++++++++++++ This is a regular weekly summary of FreeBSD's cutting-edge development. It is intended to help the FreeBSD community keep up with the fast-paced work going on in FreeBSD-CURRENT by distilling the deluge of data from the CVS mailing list into a (hopefully) easy-to-read newsletter. This newsletter is marked up in reStructuredText_, so any odd punctuation that you see is likely intended for the reST parser. .. _reStructuredText: http://docutils.sourceforge.net/rst.html You can get old summaries, and an HTML version of this one, at http://www.xl0.org/FreeBSD/. Please send any comments to Mark Johnston (mark at xl0.org). If you would like to get the summary without subscribing to current_at_, please send mail to freebsd-cvs-summary-subscribe_at_lists.enderunix.org. Thanks to Omer Faruk Sen and EnderUNIX for hosting this list. For Lukasz Dudek and Szymon Roczniak's Polish translations of these summaries, which may lag the English ones slightly, please see http://mocart.pinco.pl/FreeBSD/. .. contents:: ============ New features ============ tail can now follow multiple files ---------------------------------- Paul Richards (paul) added support to tail, which shows the last few lines of a file, to follow (-f) more than one file. This means that you can use tail -f file1 file2 file3 to show any added lines to any of the files. http://www.freebsd.org/cgi/mid.cgi?200411041918.iA4JIJsd028943 Device vnode bypass added ------------------------- Poul-Henning Kamp (phk) added support for a new feature, called device vnode bypass. When this bypass is enabled, it allows device access to go directly to the DEVFS device layer instead of having to pass through another layer (the vnode layer) first. Drivers that do not need the Giant system lock to run can now read and write without it; this includes /dev/null, /dev/zero, and hard disks. Poul-Henning gives the following example:: dd if=/dev/zero of=/dev/null bs=1 count=1000000 which, on a 700MHz K7 machine, runs twice as fast with the bypass enabled. The new code is disabled by default; to enable it, set the sysctl vfs.devfs.fops to 1 (vfs.devfs.fops=1) in /boot/loader.conf and reboot. http://www.freebsd.org/cgi/mid.cgi?200411081046.iA8AklqT003808 USB 2.0 (EHCI) enhancements --------------------------- Ian Dowse (iedowse) merged numerous fixes for USB 2 from NetBSD. The changes improve standards compliance and improve the driver's status messages, among other bug fixes. http://www.freebsd.org/cgi/mid.cgi?200411030152.iA31qoLv069164 =============== Notable changes =============== Default i386 and ia64 HZ setting changed to 1000 ------------------------------------------------ Poul-Henning Kamp (phk) changed the i386 default setting of the HZ system variable, which controls the granularity of the system clock, from 100 to 1000. This means that the standard system granularity will be 1 ms instead of 10 ms. In the same thread, Max Laier (mlaier) gave a link to `discussion of the change`_. Nate Lawson (njl) warned of a low hum, audible on some laptops and caused by ACPI's CPU idling, that is more noticeable at 1000 HZ than 100. Marcel Moolenaar (marcel) later changed the HZ on ia64 to 1000 as well. .. _`discussion of the change`: http://lists.freebsd.org/pipermail/freebsd-arch/2004-November/003085.html http://www.freebsd.org/cgi/mid.cgi?200411061133.iA6BXhHT086940 http://www.freebsd.org/cgi/mid.cgi?200411080450.iA84o2Te081727 T/TCP support removed from the system ------------------------------------- Andre Oppermann (andre) removed support for T/TCP (Transactional TCP), as defined in `RFC 1644`_, from the network stack. He gives a link to `a thread discussing the reasons`_ for the change. He specifically mentions that its code makes the TCP code very difficult to maintain, that it is unsupported on OSes other than FreeBSD, and that the design is full of security flaws. Andre also notes that it is disabled by default in FreeBSD and that it has different API calls than TCP, so it is not supported by any common network applications. .. _`RFC 1644`: http://www.rfc-editor.org/rfc/rfc1644.txt .. _`a thread discussing the reasons`: http://www.freebsd.org/cgi/mid.cgi?4177C8AD.6060706 http://www.freebsd.org/cgi/mid.cgi?200411022222.iA2MMMJs054892 gx driver removed ----------------- Poul-Henning Kamp (phk) removed the gx driver, which used to support Intel Pro/1000 Gigabit Ethernet cards. All of the hardware it supported is now supported by the em driver, which is more actively maintained. http://www.freebsd.org/cgi/mid.cgi?200411081905.iA8J5MDt044879 ================= Discussion topics ================= Moving boot messages around --------------------------- Poul-Henning Kamp (phk) made a change to the em driver, for some Intel network cards, so that the "Link is up" and "Link is down" status messages would be shown only when the system is started with "boot -v". Scott Long (scottl) replied to this, saying, "I think that in general we can start eyeing a lot of similar verbosity in all of our drivers, both at boot and at runtime. For example most nic drivers print out their MAC and all of their possible negotiation rates on attach. This same info can be obtained from ifconfig. Why clutter the boot with it?" Luigi Rizzo (rizzo) replied, "because sometimes you might need to enable some MAC-based filter before you have a chance to access the box. e.g. to set dhcp/bootp anbd the like. [ . . . ] I think the link-up/link-down thing is different because it happens not just at boot but also whenever the link status changes [ . . . ], and this is annoying on the console." Scott replied, "I'm not saying that these things should be removed, just places under bootverbose. Situations like what you are saying are rare and/or one-time occurances. And if you look at drivers like if_ath(no offense meant towards Sam):: ath0: <Atheros 5212> mem 0xf4010000-0xf401ffff irq 11 at device 0.0 on cardbus1 ath0: mac 5.6 phy 4.1 5ghz radio 1.7 2ghz radio 2.3 ath0: Ethernet address: 00:0c:41:15:5f:94 ath0: 11a rates: 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps ath0: 11b rates: 1Mbps 2Mbps 5.5Mbps 11Mbps ath0: 11g rates: 1Mbps 2Mbps 5.5Mbps 11Mbps 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps ath0: turbo rates: 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps The last 4 lines could easily go under bootverbose. I'd even wager that the second line can be hidden also[.]" Luigi responded, "ok, but that one time is precisely when you have little or no chance to intervene. Anyways, as long as you keep the MAC printout on by default i am happy :)" M. Warner Losh (imp) also replied to Scott's example, saying, "Agreed. The second line should likely be exported as a series of sysctls [ . . . ] if they are important." John Baldwin (jhb) replied to Scott too, saying, "I think bootverbose needs to be split up some." He outlined a scheme that would create different categories for boot messages that could be enabled independently. Poul-Henning answered, "I think that would be overkill." Bjoern A. Zeeb responded, "I think it's overkill to enable all verbose msgs just to see link state messages in my logs again [ . . . ]." Poul-Henning replied, "Link states are visible with ifconfig(8)." Bjoern replied, "of course they are but who will notify me or write them to my logs so I have timestamps ? Do I need to write a small daemon to regularly and fast enough poll for link state changes now ?" Poul-Henning said, "I'm not against logging link states, but if we do it, we should not arbitrarily limit it to when the link is gigE." John-Mark Gurney (jmg) responded, "Sounds like we need to expand devd for this.." Andre Oppermann (andre) pointed out, "The network drivers poll their PHY via MII every second for connection status. You don't have to worry about milliseconds or so." Robert Watson (rwatson) replied to Bjoern's last post too, saying, "If the information provided by "route monitor" isn't sufficiently useful and detailed (i.e., easily parsable) we should make it more useful, I think." Garrett Wollman (wollman) suggested, "Why can't we simply log() a message?" Returning to the ath driver example, Sam Leffler (sam) followed up, "The 2nd line cannot be removed. I'm ok with moving the rates under bootverbose." Gleb Smirnoff (glebius) also replied regarding the link up/down messages, saying, "[M]y vote is: keep initialization printfs as is, remove any runtime printfs." http://www.freebsd.org/cgi/mid.cgi?200411031411.iA3EBIqL012134 ================= Committer changes ================= Xin Li has graduated to full committerhood and no longer needs his mentor's approval for commits. http://www.freebsd.org/cgi/mid.cgi?200411031936.iA3JacQT030309 Philip Paeps (philip) has also reached full committerhood and is no longer being mentored. http://www.freebsd.org/cgi/mid.cgi?200411070628.iA76SZjw061430 =================== Important bug fixes =================== Potential denial of service in syslog fixed ------------------------------------------- Gleb Smirnoff (glebius) fixed a problem in the syslog subsystem, which collects log messages and writes them to various files, that could result in a denial of service if a local user floods /var/run/log. If a user was able to flood syslog with messages, it could cause programs like sshd and su to run into infinite loops when trying to log. Syslog will now create a second logging socket, /var/run/logpriv, to receive privileged messages, and log clients will no longer loop forever on this socket. http://www.freebsd.org/cgi/mid.cgi?200411042309.iA4N9v06043345 =============== Other bug fixes =============== Yar Tikhiy (yar) fixed a bug in the FTP server that could cause it to log incorrect filenames. The bug was noticed by Nick Leuta. http://www.freebsd.org/cgi/mid.cgi?200411030652.iA36qeOO084573 Max Laier (mlaier) fixed a bug in PF, the OpenBSD packet filter, that was preventing it from using renamed interfaces. The bug was reported by Ed Schouten as `PR 72444`_. .. _`PR 72444`: http://www.freebsd.org/cgi/query-pr.cgi?pr=72444 http://www.freebsd.org/cgi/mid.cgi?200411031721.iA3HLD72025259 Ceri Davies (ceri) committed a fix to ancontrol, the control program for Aironet wireless network cards, for a bug that caused it to reject the -f (set fragmentation threshhold) argument. The notification and fix were submitted by Irina Liakh in `PR 73229`_. .. _`PR 73229`: http://www.freebsd.org/cgi/query-pr.cgi?pr=73229 http://www.freebsd.org/cgi/mid.cgi?200411032045.iA3KjqsC032868Received on Mon Nov 08 2004 - 21:46:42 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:21 UTC