re: problem with new kernel

re: problem with new kernel

[eclark]

Previous, I was using a RH stock kernel. This was rebuilt to be monolithic 
with allthe required modules. However, now any binary which uses 
set_thread_area segfaults. I know this is a kernel issue, as other kernels I 
have used do not have this same issue. I am running 2.4.33-pre. Below is an 
example strace of nslookup, one of the relevant binaries now segfaulting.
Above the big strace (total strace of nslookup) is where I believe the problem 
is coming from. Any help would be appreciated, as I have no clue what went 
wrong with this kernel build.

---

set_thread_area({entry_number:-1 -> -1, base_addr:0x4044ec60, limit:1048575, 
seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1, 
seg_not_present:0, useable:1}) = -1 ENOSYS (Function not implemented)
modify_ldt(1, {entry_number:0, base_addr:0x4044ec60, limit:1048575, 
seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1, 
seg_not_present:0, useable:1}, 16) = 0
---

 [root@dev linux-2.4.32]# strace /usr/bin/nslookup foo.com
execve("/usr/bin/nslookup", ["/usr/bin/nslookup", "foo.com"], [/* 20 vars */]) 
= 0
uname({sys="Linux", node="dev.test.com", ...}) = 0
brk(0)                                  = 0x80566c8
open("/etc/ld.so.preload", O_RDONLY)    = -1 ENOENT (No such file or 
directory)
open("/etc/ld.so.cache", O_RDONLY)      = 3
fstat64(3, {st_mode=S_IFREG|0644, st_size=43271, ...}) = 0
old_mmap(NULL, 43271, PROT_READ, MAP_PRIVATE, 3, 0) = 0x40016000
close(3)                                = 0
open("/usr/lib/libdns.so.16", O_RDONLY) = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\320\"\1"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=1026680, ...}) = 0
old_mmap(NULL, 1030972, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x40021000
old_mmap(0x40119000, 16384, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0xf7000) = 0x40119000
close(3)                                = 0
open("/usr/lib/libisc.so.7", O_RDONLY)  = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\200z\0"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=221592, ...}) = 0
old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 
0x4011d000
old_mmap(NULL, 224824, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x4011e000
old_mmap(0x40154000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0x35000) = 0x40154000
close(3)                                = 0
open("/lib/libcrypto.so.4", O_RDONLY)   = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\360\252"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=976284, ...}) = 0
old_mmap(NULL, 989496, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x40155000
old_mmap(0x40232000, 73728, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0xdd000) = 0x40232000
old_mmap(0x40244000, 10552, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|
MAP_ANONYMOUS, -1, 0) = 0x40244000
close(3)                                = 0
open("/lib/libnsl.so.1", O_RDONLY)      = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0 <\0\000"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=91040, ...}) = 0
old_mmap(NULL, 84864, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x40247000
old_mmap(0x40259000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0x11000) = 0x40259000
old_mmap(0x4025a000, 7040, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|
MAP_ANONYMOUS, -1, 0) = 0x4025a000
close(3)                                = 0
open("/lib/tls/libpthread.so.0", O_RDONLY) = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\0G\0\000"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=98004, ...}) = 0
old_mmap(NULL, 65140, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x4025c000
old_mmap(0x40269000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0xc000) = 0x40269000
old_mmap(0x4026a000, 7796, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|
MAP_ANONYMOUS, -1, 0) = 0x4026a000
close(3)                                = 0
open("/lib/tls/libc.so.6", O_RDONLY)    = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\200X\1"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=1572440, ...}) = 0
old_mmap(NULL, 1279916, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x4026c000
old_mmap(0x4039f000, 12288, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0x132000) = 0x4039f000
old_mmap(0x403a2000, 10156, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|
MAP_ANONYMOUS, -1, 0) = 0x403a2000
close(3)                                = 0
open("/usr/kerberos/lib/libgssapi_krb5.so.2", O_RDONLY) = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\340H\0"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=76712, ...}) = 0
old_mmap(NULL, 75588, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x403a5000
old_mmap(0x403b7000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0x12000) = 0x403b7000
close(3)                                = 0
open("/usr/kerberos/lib/libkrb5.so.3", O_RDONLY) = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\320\362"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=385252, ...}) = 0
old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 
0x403b8000
old_mmap(NULL, 384636, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x403b9000
old_mmap(0x40415000, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0x5c000) = 0x40415000
close(3)                                = 0
open("/usr/kerberos/lib/libcom_err.so.3", O_RDONLY) = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\240\t\0"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=5540, ...}) = 0
old_mmap(NULL, 4520, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x40417000
old_mmap(0x40418000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0x1000) = 0x40418000
close(3)                                = 0
open("/usr/kerberos/lib/libk5crypto.so.3", O_RDONLY) = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\0&\0\000"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=72552, ...}) = 0
old_mmap(NULL, 73108, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x40419000
old_mmap(0x4042a000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0x11000) = 0x4042a000
close(3)                                = 0
open("/lib/libresolv.so.2", O_RDONLY)   = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\20*\0\000"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=76540, ...}) = 0
old_mmap(NULL, 73604, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x4042b000
old_mmap(0x4043a000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0xf000) = 0x4043a000
old_mmap(0x4043b000, 8068, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED|
MAP_ANONYMOUS, -1, 0) = 0x4043b000
close(3)                                = 0
open("/lib/libdl.so.2", O_RDONLY)       = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\260\32"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=14868, ...}) = 0
old_mmap(NULL, 12244, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x4043d000
old_mmap(0x4043f000, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0x1000) = 0x4043f000
close(3)                                = 0
open("/usr/lib/libz.so.1", O_RDONLY)    = 3
read(3, "\177ELF\1\1\1\0\0\0\0\0\0\0\0\0\3\0\3\0\1\0\0\0\340\31"..., 512) = 
512
fstat64(3, {st_mode=S_IFREG|0755, st_size=52584, ...}) = 0
old_mmap(NULL, 55564, PROT_READ|PROT_EXEC, MAP_PRIVATE, 3, 0) = 0x40440000
old_mmap(0x4044c000, 8192, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_FIXED, 3, 
0xb000) = 0x4044c000
close(3)                                = 0
old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 
0x4044e000
old_mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 
0x4044f000
set_thread_area({entry_number:-1 -> -1, base_addr:0x4044ec60, limit:1048575, 
seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1, 
seg_not_present:0, useable:1}) = -1 ENOSYS (Function not implemented)
modify_ldt(1, {entry_number:0, base_addr:0x4044ec60, limit:1048575, 
seg_32bit:1, contents:0, read_exec_only:0, limit_in_pages:1, 
seg_not_present:0, useable:1}, 16) = 0
munmap(0x40016000, 43271)               = 0
set_tid_address(0x4044eca8)             = -1 ENOSYS (Function not implemented)
rt_sigaction(SIGRTMIN, {0x40260660, [], SA_RESTORER|SA_SIGINFO, 0x40266f80}, 
NULL, 8) = 0
rt_sigprocmask(SIG_UNBLOCK, [RTMIN], NULL, 8) = 0
getrlimit(RLIMIT_STACK, {rlim_cur=8192*1024, rlim_max=RLIM_INFINITY}) = 0
rt_sigaction(SIGINT, {0x4013da50, ~[RTMIN], SA_RESTORER, 0x40266f88}, NULL, 8) 
= 0
rt_sigaction(SIGTERM, {0x4013da50, ~[RTMIN], SA_RESTORER, 0x40266f88}, NULL, 
8) = 0
rt_sigaction(SIGPIPE, {SIG_IGN}, NULL, 8) = 0
rt_sigaction(SIGHUP, {SIG_DFL}, NULL, 8) = 0
rt_sigprocmask(SIG_BLOCK, [HUP INT TERM], NULL, 8) = 0
socket(PF_INET, SOCK_STREAM, IPPROTO_IP) = 3
close(3)                                = 0
socket(PF_INET6, SOCK_STREAM, IPPROTO_IP) = -1 EAFNOSUPPORT (Address family 
not supported by protocol)
futex(0x401546b0, FUTEX_WAKE, 2147483647) = -1 ENOSYS (Function not 
implemented)
brk(0)                                  = 0x80566c8
brk(0x80776c8)                          = 0x80776c8
brk(0)                                  = 0x80776c8
brk(0x8078000)                          = 0x8078000
mmap2(NULL, 8392704, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 
0x40450000
mprotect(0x40450000, 4096, PROT_NONE)   = 0
clone(child_stack=0x40c50b04, flags=CLONE_VM|CLONE_FS|CLONE_FILES|
CLONE_SIGHAND|CLONE_THREAD|CLONE_SYSVSEM|CLONE_SETTLS|CLONE_PARENT_SETTID|
CLONE_CHILD_CLEARTID|CLONE_DETACHED, parent_tidptr=0x40c50bf8, 
{entry_number:0, base_addr:0x40c50bb0, limit:1048575, seg_32bit:1, 
contents:0, read_exec_only:0, limit_in_pages:1, seg_not_present:0, 
useable:1}, child_tidptr=0x40c50bf8) = 5404
--- SIGSEGV (Segmentation fault) @ 0 (0) ---
+++ killed by SIGSEGV +++

re: problem with new kernel

[Arjan van de Ven]

On Fri, 2006-06-30 at 10:46 -0400, eclark wrote:
> Previous, I was using a RH stock kernel. This was rebuilt to be monolithic 
> with allthe required modules. However, now any binary which uses 
> set_thread_area segfaults. I know this is a kernel issue, as other kernels I 
> have used do not have this same issue. I am running 2.4.33-pre. Below is an 
> example strace of nslookup, one of the relevant binaries now segfaulting.
> Above the big strace (total strace of nslookup) is where I believe the problem 
> is coming from. Any help would be appreciated, as I have no clue what went 
> wrong with this kernel build.

Hi,

you're running a kernel without NPTL support on a distribution that
apparently expects NPTL support to be in the kernel... the failure mode
isn't nice but failure at all isn't totally unexpected...... NPTL is
needed for certain functionality and if a distribution expects that to
be there.. things may well go very wonky if absent. (yes glibc tries to
emulate this but the emulation is quite limited and not really possible)

One example of such thing is cross-process mutexes and other locks; NPTL
allows that (via futexes on shared pages), and things like rpm use this
extensively. Another example is thread local storage; emulation is
fragile via LDT's and such.

Greetings,
   Arjan van de Ven

Re: problem with new kernel

[eclark]

Excellent! How do I get NPTL built into the kernel. I had already been 
tohttp://people.redhat.com/~drepper/ this morning after googling my straces 
for a few hours, but nothing that indicated what needed to be done to get it 
into the kernel. Any assistance would be superb!
  
On Friday 30 June 2006 12:45 pm, Arjan van de Ven wrote:
> On Fri, 2006-06-30 at 10:46 -0400, eclark wrote:
> > Previous, I was using a RH stock kernel. This was rebuilt to be
> > monolithic with allthe required modules. However, now any binary which
> > uses set_thread_area segfaults. I know this is a kernel issue, as other
> > kernels I have used do not have this same issue. I am running 2.4.33-pre.
> > Below is an example strace of nslookup, one of the relevant binaries now
> > segfaulting. Above the big strace (total strace of nslookup) is where I
> > believe the problem is coming from. Any help would be appreciated, as I
> > have no clue what went wrong with this kernel build.
>
> Hi,
>
> you're running a kernel without NPTL support on a distribution that
> apparently expects NPTL support to be in the kernel... the failure mode
> isn't nice but failure at all isn't totally unexpected...... NPTL is
> needed for certain functionality and if a distribution expects that to
> be there.. things may well go very wonky if absent. (yes glibc tries to
> emulate this but the emulation is quite limited and not really possible)
>
> One example of such thing is cross-process mutexes and other locks; NPTL
> allows that (via futexes on shared pages), and things like rpm use this
> extensively. Another example is thread local storage; emulation is
> fragile via LDT's and such.
>
> Greetings,
>    Arjan van de Ven

Re: problem with new kernel

[eclark]

My apologies for the double email. For what its worth:

[root@dev linux-2.4.32]# getconf GNU_LIBPTHREAD_VERSION
NPTL 0.60

Not sure what that implies (other than its still broke. :| )

On Friday 30 June 2006 12:45 pm, Arjan van de Ven wrote:
> Hi,
>
> you're running a kernel without NPTL support on a distribution that
> apparently expects NPTL support to be in the kernel... the failure mode
> isn't nice but failure at all isn't totally unexpected...... NPTL is
> needed for certain functionality and if a distribution expects that to
> be there.. things may well go very wonky if absent. (yes glibc tries to
> emulate this but the emulation is quite limited and not really possible)

Re: problem with new kernel

[Arjan van de Ven]

On Fri, 2006-06-30 at 13:10 -0400, eclark wrote:
> Excellent! How do I get NPTL built into the kernel.

realistically NPTL on 2.4 is a real big pain in the <>. So I would
suggest to either use a distribution kernel with NPTL patched in, or use
a 2.6 kernel. Making your own NPTL patch for 2.4.. lets say it would
take a really really experienced kernel hacker at least 3 weeks... and
stability would be.. brittle. Just don't go there ;(

Greetings,
   Arjan van de Ven