On 01/12/2018 04:20 AM, Vladimir Sementsov-Ogievskiy wrote: > 11.01.2018 02:08, Eric Blake wrote: >> Rather than making every callsite perform length sanity checks >> and error reporting, add the helper functions nbd_opt_read() >> and nbd_opt_drop() that use the length stored in the client >> struct; also add an assertion that optlen is reduced to zero >> after each option is handled. >> >> Note that the call in nbd_negotiate_handle_export_name() does >> not use the new helper (in part because the server cannot >> reply to NBD_OPT_EXPORT_NAME - it either succeeds or the >> connection drops). >>
>>
>> +/* Drop remainder of the current option, after sending a reply with
>
> looks a bit weird: actually you drop the remainder _before_ sending a
> reply)
Good catch. I'll fix it with s/after/and/
>
>> + * the given error type and message. Return -errno on read or write
>
> also, unrelated note, -errno is always forced to -EIO, because of
> nbd_read realization.
> and this note applies to many other places here. It is correct (EIO is
> errno, why not?),
> but it may be not bad to note it somewhere..
Someday nbd_read() might fail with something other than EIO (ESHUTDOWN,
perhaps?), in which case leaving this documented as -errno would be
better than hardcoding that EIO is the only failure for now.
>> @@ -812,14 +819,9 @@ static int nbd_negotiate_options(NBDClient
>> *client, uint16_t myflags,
>> break;
>>
>> default:
>> - if (nbd_drop(client->ioc, length, errp) < 0) {
>> - return -EIO;
>> - }
>> - ret = nbd_negotiate_send_rep_err(client,
>> - NBD_REP_ERR_UNSUP,
>> errp,
>> - "Unsupported option
>> 0x%"
>> - PRIx32 " (%s)", option,
>> -
>> nbd_opt_lookup(option));
>> + ret = nbd_opt_drop(client, NBD_REP_ERR_UNSUP, errp,
>> + "Unsupported option 0x%" PRIx32 "
>> (%s)",
>> + option, nbd_opt_lookup(option));
>> break;
>> }
>> } else {
>> @@ -842,6 +844,7 @@ static int nbd_negotiate_options(NBDClient
>> *client, uint16_t myflags,
>> if (ret < 0) {
>> return ret;
>> }
>> + assert(!client->optlen);
>
> isn't it from 2/6?
No, this is a second instance of the assertion, the one that applies
between each option (which I couldn't do in 2/6 because not all options
were manipulating optlen back then). Maybe I can tweak the commit
messages to make that more obvious.
>
>> }
>> }
>>
>
> anyway,
>
> Reviewed-by: Vladimir Sementsov-Ogievskiy <[email protected]>
Thanks for the careful attention to detail.
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org
signature.asc
Description: OpenPGP digital signature
