Ok, I think I have a hint. I ran strace on qmail-smtpd on a working server,
and I get:
select(1, [0], NULL, NULL, {60, 0}starttls
) = 1 (in [0], left {34, 700000})
read(0, "starttls\n", 1024) = 9
open("control/servercert.pem", O_RDONLY) = 3
fstat64(3, {st_mode=S_IFREG|0644, st_size=2141, ...}) = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) =
0xb7f01000
read(3, "-----BEGIN RSA PRIVATE KEY-----\n"..., 4096) = 2141
I think it's choking ok the read statement, not the mmap2. When it reads
file handle 3, it's showing what it gets (BEGIN RSA PRIVATE KEY).
>From previous strace output, I see that qmail-smtpd is using file handle 3
for various things. At this point though, it's pointing to servercert.pem.
Chew on this, and I'll keep comparing straces.
Chris Godwin wrote:
> Yes but the memory address changes when do a second and third strace.
>
> Chris Godwin
> Linux/Unix Consultant
> Network Logistic, Inc.
> Get help at http://www.networklogistic.com/help
>
>
> -----Original Message-----
> From: Eric "Shubes" [mailto:[EMAIL PROTECTED]
> Sent: Wednesday, October 04, 2006 4:04 PM
> To: [email protected]
> Subject: Re: [qmailtoaster] STARTTLS Segmentation Fault
>
> Ok, so be it.
>
> I'm no expert at reading an strace, but let's have a look at it again:
>
> select(1, [0], NULL, NULL, {1154, 135000}) = 1 (in [0], left {1146,
> 148000})
> read(0, "ehlo localhost\r\n", 1024) = 16
> stat64("control/servercert.pem", {st_mode=S_IFREG|0777, st_size=1689,
> ...}) = 0
> select(2, NULL, [1], NULL, {1200, 0}) = 1 (out [1], left {1200, 0})
> write(1, "250-ha.networklogistic.net - Wel"..., 170) = 170
> select(1, [0], NULL, NULL, {1200, 0}) = 1 (in [0], left {1193,
> 949000})
> read(0, "starttls\r\n", 1024) = 10
> brk(0) = 0x9cf4000
> brk(0x9d15000) = 0x9cf4000
> mmap2(NULL, 1048576, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS,
> -1, 0)
> = 0xb7ee8000
> --- SIGSEGV (Segmentation fault) @ 0 (0) ---
>
> It's choking on the mmap2 function, yes?
>
> Chris Godwin wrote:
>> Well actually I have to change chkuser in order to disallow sender_mx
>> checking. This is to enable email from hosts on our network that send
>> out cron emails but have no mx records.
>>
>> Chris Godwin
>> Linux/Unix Consultant
>> Network Logistic, Inc.
>> Get help at http://www.networklogistic.com/help
>>
>>
>
--
-Eric 'shubes'
---------------------------------------------------------------------
QmailToaster hosted by: VR Hosted <http://www.vr.org>
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
