1 |
Nathan, thanks so much for all the effort you've put into this. I am |
2 |
quite relieved we now know more about it and that it's no longer a |
3 |
complete mystery. |
4 |
|
5 |
As a data point, I have the issue (my "test site" that is unreachable is |
6 |
ftp.astron.com), and I am on a Fujitsu Lifebook P7120D, which has this |
7 |
processor: |
8 |
|
9 |
* Intel Pentium M Processor Ultra Low Voltage 753 (1.20 GHz, 2 MB L2 |
10 |
cache, 400 MHz FSB) |
11 |
|
12 |
-Joe |
13 |
|
14 |
|
15 |
Nathan Smith wrote: |
16 |
> On 4/20/07, Timothy Redaelli <drizzt@g.o> wrote: |
17 |
>> Can you try to recompile the kernel with gcc 4 but with -O0? |
18 |
>> cd /usr/src/sys/i386/compile/GENERIC ; COPTFLAGS=-O0 make |
19 |
> |
20 |
> I can confirm that compiling with gcc-4 and COPTFLAGS=-O0 produces a |
21 |
> working kernel (i.e. no malformed tcp packets). This was after I |
22 |
> found that the gcc41 patch is also not causing the problem. After |
23 |
> getting a working kernel with gcc-4 I went back and used the default |
24 |
> COPTFLAGS again and it broke again. So I think this gives us a fair |
25 |
> amount of certainty that the problem is caused by gcc-4 optimizations. |
26 |
> |
27 |
> As to the reason why some FreeBSD users experience this and other do |
28 |
> not, I am not sure. The symptoms (broken TCP traffic to some sites) |
29 |
> depend entirely upon the sites you are connecting to and how your |
30 |
> connections are routed. So it is possible that somebody could have |
31 |
> this problem by not realize it. |
32 |
> |
33 |
> However, using tcpdump I can tell that some machines apparently do not |
34 |
> experience this problem, so it may be processor-specific (which makes |
35 |
> sense given that it is an optimization problem). I use a Pentium M. |
36 |
> What are other people using that have been affected by the problem? |
37 |
> |
38 |
-- |
39 |
gentoo-bsd@g.o mailing list |