On 05/06/2017 03:25, Sage Weil wrote: > On Sun, 4 Jun 2017, Willem Jan Withagen wrote: >> On 30/05/2017 03:30, Sage Weil wrote: >>> It does. There is a helper somewhere that maps error codes in MOSDOpReply >>> that go over the wire that was added for an AIX port a year or two ago. >>> look for errorcode32_t and ceph_to_host_errno etc >> >> 'mmm >> >> It looks like it is the other way around as what I'm looking for... >> These are the linux errno's translated to AIX/Darwin codes. So the >> native OS does not get upset when playing client. These .cc files are >> not even compiled in the standard ceph tree, as far as I can see. >> >> So I guess what I'm looking for is to issue the correct errorcodes on >> the line when also acting as server. And as far as I can imagine will >> need both: ceph_to_host_errno() and host_to_ceph_errno() and start >> glueing that in. >> Now the question is: I that restricted to one location to insert as >> well, or go I need to go chasing all places where the errcode is de/encoded. > > 1. Write the ceph_to_host_errno and host_to_ceph_errno methods > 2. Fix the errorcode32_t to use them for encode *and* decode. > 3. Identify the handful of places where we send errcodes over the wire > (mostly MOSDOpReply and friends) and switch the type to errorcode32_t so > that it is translated during encode/decode. > > Happy to help with 2 and 3 once we have 1 in place! Great, Started working on 1, only to realise how disjunct they are. I guess 3 will be the hardest not to miss any spots. As our Austrian-American friend says: I'll be back. --WjW >>> On Mon, 29 May 2017, Willem Jan Withagen wrote: >>> >>>> Hi Sage >>>> >>>> I came accros the following, dataing september 2011, writen by you: >>>> ==== >>>> These error codes pass over the wire, so we should make sure the actual >>>> values are consistent across the platforms, so a FreeBSD client can talk >>>> to a Linux server, or vice-versa. >>>> >>>> I'm not sure if that's the case for other error codes...? >>>> ==== >>>> >>>> Now the trick question is: Does ENODATA travel over the wire? >>>> >>>> Because the way I designed it now >>>> ENOATTR on FreeBSD is 87 >>>> ENODATA on Linux is 61 >>>> >>>> So that is not going to match, if it is communicated. >>>> >>>> --WjW >>>> >>>> >> >> -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
