Hi Henrik, On Thu, 2013-12-05 at 15:37 +0100, Henrik Goldman wrote: > Hello, > > We're having a setup running with Ubuntu 13.04 and 13.10 for target and > VMware 5.1/5.5 as the client. In this setup we're using 4 Gbit FC using > QLA2460 adapters. Everything works quite well except we keep on getting > kernel warnings like: > > [1401151.960377] SCSI OP 2ah with too big sectors 2056 exceeds backend > hw_max_sectors: 2048 > > This happens every few seconds and is consistent across several systems. > > Our setup is Raid 10 and 5 using mdadm using FILEIO in buffered mode. We > chose this since it gives really great performance and uses memory for > caching. > > However from what I read then this value 2048 is hardcoded into the kernel > and there is no way of changing this. > Am I right on this? > > Any way to get this fixed or some workarounds for this? > A bit of background here.. This requirement is enforced in FILEIO specifically for 4k block_size in order to prevent greater than 2k iovecs being passed into vfs_[writev,readv] logic, a limitation of the total number of iovecs enforced by logic in the VFS below FILEIO code. However with a 512 byte block_size, this effectively limits the max request size to 1MB, as you've reported above. A quick work-around here is to increase FD_MAX_SECTORS to 4096 or 8192 (2MB or 4MB max request size), as with ESX you'll only be using 512 block_size, so this will be OK. I'm considering a couple of difference approaches here for a proper solution, but changing FD_MAX_SECTOR to a larger value for 512 block_size should get you up and running. Also, I'd recommend explicitly setting the ESX side max request size to the new FD_MAX_SECTOR * block_size. --nab -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
