This patch does not introduceĪny new races as such w.r.t set_capacity()/get_capacity(). Modify set_capacity()/get_capacity() due to sheer number of call sitesĪnd I am afraid that change might break something. Theoritically race should exist there too but this patch does not There is one exception though, set_capacity()/get_capacity().
Now all the access to hd_struct->nr_sects should happen using sequenceĬounter read/update helper functions part_nr_sects_read/part_nr_sects_write.
SequenceĬounter have been used so that readers don't have to take bdev mutex lockĪs we call sector_in_part() very frequently. ThisĬan lead to issues like reading inconsistent size of a partition. Nr_sects can be non-atomic on 32bit machines with 64bit sector_t. One might extend a partition while IO is happening to it and update of This patch converts hd_struct->nr_sects into sequence counter because Resending this patch again with your Signed-off-by.Īdd a new operation code (BLKPG_RESIZE_PARTITION) to the BLKPG ioctl thatĪllows altering the size of an existing partition, even if it is currently Asking user to call two different utilities Especially a utility which changes partition details both I wrote user space part only to do testing. My original util-linux patches also added a -u update mode to kpartx, which I think is the more useful interface than the lower level resizepart command, but I suppose I can rebase it to apply on top of this patch. > -BEGIN PGP SIGNED MESSAGE- > Hash: SHA1 > On 05:34 PM, wrote: > Phillip, do let me know if I should put your signed-off-by also in the > patch. X-loop: Re: block: Support online resize ofĬontent-Type: text/plain charset="us-ascii"
Received: by (Postfix, from userid 10451) X-Original-To: from ( )īy (Postfix) with ESMTP id 9AF2DDFF34
0 kommentar(er)
