[Qemu-devel] Re: [PATCH 3/3] savevm: Convert loadvm handlers list to LIST

[Juan Quintela]

Luiz Capitulino <address@hidden> wrote:
> On Tue, 01 Sep 2009 00:08:43 +0200
> Juan Quintela <address@hidden> wrote:
>
>> Luiz Capitulino <address@hidden> wrote:
>> > On Fri, 28 Aug 2009 22:31:57 +0200
>> > Juan Quintela <address@hidden> wrote:
>> >
>> >> 
>> >> Signed-off-by: Juan Quintela <address@hidden>
>> >> ---
>> >>  savevm.c |   20 +++++++++++---------
>> >>  1 files changed, 11 insertions(+), 9 deletions(-)
>> >> 
>> >> diff --git a/savevm.c b/savevm.c
>> >> index baef277..9836c60 100644
>> >> --- a/savevm.c
>> >> +++ b/savevm.c
>> >> @@ -1260,10 +1260,10 @@ static SaveStateEntry *find_se(const char *idstr, 
>> >> int instance_id)
>> >>  }
>> >> 
>> >>  typedef struct LoadStateEntry {
>> >> +    LIST_ENTRY(LoadStateEntry) entry;
>> >>      SaveStateEntry *se;
>> >>      int section_id;
>> >>      int version_id;
>> >> -    struct LoadStateEntry *next;
>> >>  } LoadStateEntry;
>> >> 
>> >>  static int qemu_loadvm_state_v2(QEMUFile *f)
>> >> @@ -1309,7 +1309,8 @@ static int qemu_loadvm_state_v2(QEMUFile *f)
>> >> 
>> >>  int qemu_loadvm_state(QEMUFile *f)
>> >>  {
>> >> -    LoadStateEntry *first_le = NULL;
>> >> +    LIST_HEAD(, LoadStateEntry) loadvm_handlers;
>> >
>> >  You're missing the initialization here, spot this while
>> > testing staging.
>> 
>> I looked at aio.c and guess what :)  No LIST_INIT() either.
>
>  As we talked by irc, if you are referring to 'aio_handlers' it's
> global, so it will be initialized to 0 by the kernel.

My bad.  In this case, I am just a lucky man, and in all my testing, I
always got the list intialized correctly :)

Patches fixing it sent already.

Thanks a lot, Juan.

>> My understanding is that it is not needed (but it is better to add it,
>> just in case the implementation change).
>
>  I'm getting a segfault when I try to loadvm, I can write a recipe
> to reproduce if needed.
>
>> #define LIST_HEAD_INITIALIZER(head)                                     \
>>         { NULL }
>> 
>> #define LIST_INIT(head) do {                                            \
>>         (head)->lh_first = NULL;                                        \
>> } while (/*CONSTCOND*/0)
>> 
>> 
>> This should be what it does without puting it.  Perhaps we should
>> "correct" also the other users?
>
>  Yes, IMHO. Not because it's a fix, but it's good practice to use the API
> w/o making assumptions on its internals.