URL:
http://www.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.20pre10aa1.gz
http://www.us.kernel.org/pub/linux/kernel/people/andrea/kernels/v2.4/2.4.20pre10aa1/
changelog:
Only in 2.4.20pre10aa1: 00_backout-pre9-list_t-removal-1
resurrect list_t.
Only in 2.4.20pre10aa1: 00_extraversion-10
Only in 2.4.20pre8aa2: 00_extraversion-9
Only in 2.4.20pre8aa2: 00_sched-O1-aa-2.4.19rc3-4.gz
Only in 2.4.20pre10aa1: 00_sched-O1-aa-2.4.19rc3-5.gz
Only in 2.4.20pre8aa2: 00_thinkpad-1
Only in 2.4.20pre10aa1: 00_thinkpad-2.gz
Only in 2.4.20pre8aa2: 05_vm_05_zone_accounting-1
Only in 2.4.20pre10aa1: 05_vm_05_zone_accounting-2
Only in 2.4.20pre8aa2: 60_pagecache-atomic-5
Only in 2.4.20pre10aa1: 60_pagecache-atomic-6
Only in 2.4.20pre8aa2: 93_NUMAQ-5
Only in 2.4.20pre10aa1: 93_NUMAQ-6
Only in 2.4.20pre8aa2: 97_i_size-corruption-fixes-1
Only in 2.4.20pre10aa1: 97_i_size-corruption-fixes-2
Rediffed.
Only in 2.4.20pre10aa1: 00_i8k-compile-1
Should fix i8k compilation reported on l-k.
Only in 2.4.20pre10aa1: 00_ip_conntrack-ack-seq-1
Check the sequence number of the final ack of the 3-way handshake.
(from Martin Renold)
Only in 2.4.20pre10aa1: 00_ipv6-route-fix-1
Avoid losing the route. (from YOSHIFUJI Hideaki)
Only in 2.4.20pre10aa1: 00_isofs-options-elseif-1
Fix isofs option parsing. (from Chris Mason)
Only in 2.4.20pre10aa1: 00_module-locking-fix-1
Fix module locking, the get_module_info was running without
locks so it was racing and oopsing while walking the list.
The previous locking was broken as well by walking lists
and scheduling in between (releasing the big kernel lock).
The big kernel locks is now replaced by a proper semaphore.
Only in 2.4.20pre8aa2: 00_o_direct-read-overflow-write-locking-xfs-1
Merged in mainline.
Only in 2.4.20pre10aa1: 00_skb-frag-1
nr_frags initialization fix.
Only in 2.4.20pre10aa1: 00_vmalloc-leak-fix-1
Avoid leaking vmalloc metadata.
Only in 2.4.20pre8aa2: 20_sched-o1-fixes-2
Only in 2.4.20pre10aa1: 20_sched-o1-fixes-4
Fix race condition that destabilized the scheduler.
Only in 2.4.20pre10aa1: 86_x86_64-tsc-hpet-pit-1
Newer gettimeofday code (new hpet/tsc/pit combinations).
Only in 2.4.20pre8aa2: 90_ext3-commit-interval-3
Only in 2.4.20pre10aa1: 90_ext3-commit-interval-5
Use bdflush_interval() instead of 5*HZ when commit option isn't passed
to mount (the mount option will override bdflush_interval()).
Only in 2.4.20pre8aa2: 90_proc-mapped-base-3
Only in 2.4.20pre10aa1: 90_proc-mapped-base-4
Add x86-64 and s390x support (so binaries in 32bit compatibility
mode will be able to change the mapped base, the proc entry is a noop
for all 64bit tasks even if it will showup for them too [common code
cannot differentiate between 32bit and 64bit tasks, there's simply
no API for it]).
Only in 2.4.20pre10aa1: 9900_aio-11.gz
Only in 2.4.20pre8aa2: 9900_aio-9.gz
With rawio map at most KIO_MAX_SECTORS at time (or kmalloc
can even fail in the I/O routine that will try to write
the kvec).
unplug the queue asynchronously after io_submit so io_submit won't
block to start the I/O and aio-rawio will work correctly.
Only in 2.4.20pre8aa2: 9910_shm-largepage-4.gz
Only in 2.4.20pre10aa1: 9910_shm-largepage-5.gz
Fix MAX_ORDER for non PAE compiles, so largepages can be easily
preallocated at boot and the bigpages= will work also without PAE.
Enable largepages as backing store of shmfs files when
shm-use-bigpages sysctl is set to 2.
Only in 2.4.20pre8aa2: 9960_cyclone-1
Only in 2.4.20pre10aa1: 9960_cyclone-3
Fix a few bugs in the cyclone support and use it for udelay too.
Still it's not autodetected because the summit support isn't
included in mainline.
Only in 2.4.20pre10aa1: 9970_corename-1.gz
Patch from Michael Sinz to name corefiles differently (configurable via
sysctl). BTW, make sure to never use the syscall by number or it has
few chances to work correctly, sysctl by name will work fine.
http://www.sinz.org/Michael.Sinz/Linux/
The following format options are available in that string:
%P The Process ID (current->pid)
%U The UID of the process (current->uid)
%N The command name of the process (current->comm)
%H The nodename of the system (system_utsname.nodename)
%% A "%"
Andrea
On 2002.10.12 Andrea Arcangeli wrote:
[...]
>> >
>> >why would you want to remove the list_t declaration? I don't see it.
>> >list_t is just like task_t for struct task_struct etc...
>> >
>>
>> [..] The main
>> argument was that you can't pre-declare a task_t, but you can
>> with a struct task_t.
>
>so you want to remove task_t too? If yes just grep -v typedef all over
>the tree, and at least it'll be a somehow more coherent decsion ;).
>It's not that list_t forbids you to use struct list_head for
>predeclarations.
>
There is something I do not understand:
struct tst {
struct x_t *x;
}
does not need any predeclaraion, tested with gcc -Wall, 2.96, 3.0.4, 3.2.
struct tst {
x_t *x;
}
needs it, but you can predeclare a
typedef struct x_t x_t;
if do not want the full include. So I really do not know why everybody
agreed on removing it.
???
--
J.A. Magallon <[email protected]> \ Software is like sex:
werewolf.able.es \ It's better when it's free
Mandrake Linux release 9.1 (Cooker) for i586
Linux 2.4.20-pre10-jam1 (gcc 3.2 (Mandrake Linux 9.0 3.2-2mdk))
On 2002.10.12 J.A. Magallon wrote:
>
>On 2002.10.12 Andrea Arcangeli wrote:
>[...]
>
>There is something I do not understand:
>
>struct tst {
> struct x_t *x;
>}
>
Oops,
void f(struct t_t *a)
{
}
needs it, but this works:
typedef struct t_t t_t;
void f(struct t_t *a)
{
}
struct test {
struct t_t *a;
t_t *b;
};
gcc is happy.
So all what is needed is to change
struct t_t;
to
typdef struct t_t t_t;
--
J.A. Magallon <[email protected]> \ Software is like sex:
werewolf.able.es \ It's better when it's free
Mandrake Linux release 9.1 (Cooker) for i586
Linux 2.4.20-pre10-jam1 (gcc 3.2 (Mandrake Linux 9.0 3.2-2mdk))
On 2002.10.12 John Levon wrote:
>On Sat, Oct 12, 2002 at 02:43:09AM +0200, J.A. Magallon wrote:
>
>> struct t_t;
>>
>> to
>>
>> typdef struct t_t t_t;
>
>Now wonder what the point of a typedef is when you have to expose the
>underlying type anyway ;)
>
Do not wonder of anything, it is just a hack people uses to avoid
spliting all kernel headers in public and private parts, ala X.
--
J.A. Magallon <[email protected]> \ Software is like sex:
werewolf.able.es \ It's better when it's free
Mandrake Linux release 9.1 (Cooker) for i586
Linux 2.4.20-pre10-jam1 (gcc 3.2 (Mandrake Linux 9.0 3.2-2mdk))