Re: [Xen-devel] [PATCH] docs: remove wrong statement about bug in xenstore

[Juergen Gross <jgross@xxxxxxxx>]

On 24/10/16 14:06, Andrew Cooper wrote:
> On 24/10/16 12:49, Juergen Gross wrote:
>> On 24/10/16 13:41, Wei Liu wrote:
>>> On Mon, Oct 24, 2016 at 01:27:17PM +0200, Juergen Gross wrote:
>>>> docs/misc/xenstore.txt states that xenstored will use "0" as a valid
>>>> transaction id after 2^32 transactions. This is not true. Remove that
>>>> statement.
>>>>
>>>> Signed-off-by: Juergen Gross <jgross@xxxxxxxx>
>>> Can you point me to the relevant code snippet? Better still I would like
>>> to see why it is the case in commit message.
>> Sure: tools/xenstore/xenstored_transaction.c
>>
>> do_transaction_start():
>> ...
>>         /* Pick an unused transaction identifier. */
>>         do {
>>                 trans->id = conn->next_transaction_id;
>>                 exists = transaction_lookup(conn,
>> conn->next_transaction_id++);
>>         } while (!IS_ERR(exists));
>>
>> It should be noted here that conn->next_transaction_id is initialized
>> to be 0. So the error would occur for the first transaction, too.
> 
> Cxenstored isn't the only xenstored implementation, and I can't see
> anything in the Ocaml version which mitigates this issue.  Furthermore,
> because Ocaml's int is 31 bits or 63 bits, I suspect a 64bit oxenstored
> will become unusable when the transaction id hits 4 billion and an a
> truncation occurs when writing the id into the ring.  A 32bit oxenstored
> only uses half the available transaction id space, and does wrap around
> to 0.

Okay, so either oxenstored should be corrected by some ocaml capable
developer, or I can send a patch which will limit the bug statement to
oxenstored.

Such a simple to fix problem should not be just mentioned in some text
file, but it should be fixed! Leaving the text unmodified is no option
IMHO.


Juergen

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel