There is a bug in gcc-2.7.2.3. It incorrectly lays out
structure initialisers when the `name:value;' construct is used.
Here is the degenerate case:
struct struct_1 { int a; };
struct thing {
int a;
struct struct_1 b;
};
struct thing a_thing = {
// a: 0, /* Uncomment this for correct code */
b: {0},
};
Which produces:
.file "e.c"
.version "01.01"
gcc2_compiled.:
.globl a_thing
.data
.align 4
.type a_thing,@object
.size a_thing,8
a_thing:
.zero 4
.zero 4
.long 0
.ident "GCC: (GNU) 2.7.2.3"
Note the extra `.zero 4'.
As far as I can tell the rule to follow is this:
If a structure has nested structures, and if you are
initialising one of the nested structures then you
_must_ initialise all fields which precede that nested
structure.
There are five ways to fix this:
1: Remember to initialise all fields which precede
initialised structures.
2: Arrange your struct so that all nested structs come first
and remember to initialise them all.
3: Don't use gcc-2.7.2.3 (use one of the later compilers and
put up with 50% longer build times).
4: Fix gcc-2.7.2.3
5: Don't use named initialisers.
David, your recent changes to tcp_ipv4.c make 2.7.2.3-compiled
kernels fail very strangely. I used option 2 to fix it:
--- linux-2.4.0-test10-pre3/include/net/tcp.h Sat Oct 14 17:02:04 2000
+++ linux-akpm/include/net/tcp.h Sun Oct 15 22:56:38 2000
@@ -90,6 +90,24 @@
};
extern struct tcp_hashinfo {
+ rwlock_t __tcp_lhash_lock;
+ atomic_t __tcp_lhash_users;
+ wait_queue_head_t __tcp_lhash_wait;
+ spinlock_t __tcp_portalloc_lock;
+
+ /* All sockets in TCP_LISTEN state will be in here. This is the only
+ * table where wildcard'd TCP sockets can exist. Hash function here
+ * is just local port number.
+ */
+ struct sock *__tcp_listening_hash[TCP_LHTABLE_SIZE];
+
+ /*
+ * All the below members are written once at bootup and are
+ * never written again _or_ are predominantly read-access.
+ * Hence we align to a new cache line as all the preceding members
+ * are often dirty.
+ */
+
/* This is for sockets with full identity only. Sockets here will
* always be without wildcards and will have the following invariant:
*
@@ -97,8 +115,10 @@
*
* First half of the table is for sockets not in TIME_WAIT, second half
* is for TIME_WAIT sockets only.
+ *
*/
- struct tcp_ehash_bucket *__tcp_ehash;
+ struct tcp_ehash_bucket *__tcp_ehash
+ __attribute__((__aligned__(SMP_CACHE_BYTES)));
/* Ok, let's try this, I give up, we do need a local binding
* TCP hash as well as the others for fast bind/connect.
@@ -107,24 +127,6 @@
int __tcp_bhash_size;
int __tcp_ehash_size;
-
- /* All sockets in TCP_LISTEN state will be in here. This is the only
- * table where wildcard'd TCP sockets can exist. Hash function here
- * is just local port number.
- */
- struct sock *__tcp_listening_hash[TCP_LHTABLE_SIZE];
-
- /* All the above members are written once at bootup and
- * never written again _or_ are predominantly read-access.
- *
- * Now align to a new cache line as all the following members
- * are often dirty.
- */
- rwlock_t __tcp_lhash_lock
- __attribute__((__aligned__(SMP_CACHE_BYTES)));
- atomic_t __tcp_lhash_users;
- wait_queue_head_t __tcp_lhash_wait;
- spinlock_t __tcp_portalloc_lock;
} tcp_hashinfo;
#define tcp_ehash (tcp_hashinfo.__tcp_ehash)