Kernel

Jeffrey Hsu to implement TCP ABC on DragonFlyBSD

Jeffrey Hsu (a well known FreeBSD and DragonFly networking stack developer) has committed first cut of TCP ABC implementation for DragonFlyBSD. Described in RFC 3465, this is an improvement that can provide better throughput.

How to install FreeBSD without CD-ROM, Floppy nor PXE?

Today I have experienced how to install FreeBSD from a USB device - quite simple so you can do it yourself :-)

First, initialize the USB solid state disk with bsdlabel(8):

bsdlabel -wB /dev/da0

Then, initialize a file system:

newfs -O1 /dev/da0a

Third, copy CDROM’s /boot to it.

And you can now boot from the USB disk.

3 security advisories, and FreeBSD 5.4-RELEASE has been tagged

Today, Colin has committed three security fixes that addresses three medium-risk security issues in FreeBSD kernel. That results in p11, p12 and p13 of FreeBSD 5.3-RELEASE, and has been committed to RELENG_5_4 as well so we will get that in the final release build.

Ken Smith has announced the tagging (with a mini heads-up in developers@), and then committed the naming (5.4-RELEASE) to the RELENG_5_4 tree. After that, RELENG_5_4 was tagged as RELENG_5_4_0_RELEASE, which is the last step of the release engineering that is related to the src/tree.

Please note that, this is not the final step of the final release. FreeBSD is very careful about the quality (not only as opposed to the kernel with colorful history of its security record, which is released under GPL and often release from time to time, haha, you know which is that) of a release, therefore, tagging does not mean that “we made it”, that is just a stage. Please be patient to wait for the PGP signed 5.4-RELEASE announcement BEFORE you download isos, even they were found on any sites.

Bug with sched_lend_prio?

I have enabled PREEMPTION, FULL_PREEMPTION with ULE. Now I get strange result here:

• propagate_priority gets a TD_ON_LOCK(td)
• It tries to panic(), unfortunatelly, ts->ts_lockobj here is NULL
• A fatal trap 12 happens.

It seems that ts->ts_lockobj was touched by sched_lend_prio, for unknown reasons. Maybe there is some lock releases?

老了一刚 :'(

调试能力明显下降了，本来早就应该想到的，结果整整花了一个钟头才找到故障点，老了，明明看到堆栈毁了的居然还去相信调试器：（

Charlie died due to strange kernel bug, investigating

Finally I caught the kernel panic caused by some strange bugs found in thread scheduling. Investigating the crashdump.

Proposed new RELENG_5_3 errata candidate

The following change is considered as a new errata candadite from my view:

peter 2005-04-18 21:36:17 UTC FreeBSD src repository

Modified files: (Branch: RELENG_5_4)
sys/i386/i386 genassym.c vm86bios.s
sys/i386/include pcb.h
Log:
MFC: stop i386_vm86() from either panicing or trashing random pages.

This can improve stability, however we can not easily prove that it is producable :-) So I have turned that to re@.

Peter Wemm to change FreeBSD/i386 segment limit to 4GB

To quote him:

DragonFlyBSD on SATA: Not so success

So I finally tried DragonFly on one of our test servers (not my desktop :-). It seems that DragonFly still have some issues with the SATA devices. I am trying to fix it, maybe by porting some good FreeBSD ATAng code.

DragonFly has gave us some good examples that how a user-friendly operating system should, and some different way on designing a MP-capable architecture. Thanks, Matt!

FreeBSD snapshotting/background fsck improvement idea

1. We should not caclucate the superblock summary, since it can at most lag 30 seconds’ change, and doing the caclucate is painful delay. What we should do is to transfer the responsibility to background fsck
2. newfs(8) should warn about the snapshot is actually unable to create, in case fssize is greater than maxfilesize.

The patch for 1) is available at http://people.freebsd.org/~delphij/ffs/patch-sbsummary. I will work on 2) shortly, as it’s quite simpler than 1).

• ← 上一页
• 下一页 →