Thanks for the reply. > why minimum?
> why is that? This was discussed in the previous version (https://www.spinics.net/lists/linux-virtualization/msg58232.html). As far as I know, the Linux kernel uses the same "max segments" value for a discard and a secure erase command. In the first version, I ignored the max_secure_erase_seg and secure_erase_sector_alignment config fields (just like max_write_zeroes_seg and write_zeroes_may_unmap are ignored in the write zeros command implementation). It was suggested to use the minimum "max segments" value if both VIRTIO_BLK_F_SECURE_ERASE and VIRTIO_BLK_F_DISCARD are negotiated. The same is true for the sector alignment values. > is this logic repeating code from below? I'm not sure what you mean. The idea is: At this point, the VIRTIO_BLK_F_DISCARD fields were read from the virtio config (if VIRTIO_BLK_F_DISCARD is negotiated). If max_discard_segs is 0, VIRTIO_BLK_F_DISCARD is not negotiated (or set to 0), so we should use the max_secure_erase_seg value as max_discard_segs. > Always? What's going on here? > which versions handled max_secure_erase_seg == 0? This comment is from the VIRTIO_BLK_F_DISCARD implementation. I added the max_secure_erase_seg part since I could not find how to handle the case when max_secure_erase_seg is 0 in the spec. So, like with the VIRTIO_BLK_F_DISCARD implementation, I'm setting the value to sg_elems. _______________________________________________ Virtualization mailing list [email protected] https://lists.linuxfoundation.org/mailman/listinfo/virtualization
