@raquini ah, I do see that you mentioned you were continuing updates through 6.18. I would advise against dropping functionality for a single minor release, however.
/unblock
/block
@raquini you are missing commit 653d7825c1499 ("dcssblk: mark DAX broken, remove FS_DAX_LIMITED support"), which is part of the "fs/dax: Fix ZONE_DEVICE page reference counts" patch series.
I do not know how widely used that feature is, but we don't typically remove features in minor releases.
Upstream went on to add ZONE_DEVICE support to s390, and then restored the dcssblk DAX support in commit 5b27dfb1d7b59 ("s390/dcssblk: Add DAX support").
As I see it, we have 3 options:
block layer: update with v6.19
JIRA: https://issues.redhat.com/browse/RHEL-144763
Omitted-fix: ce8ee8583ed8 ("block: use trylock to avoid lockdep circular dependency in sysfs")
Omitted-fix: dd677d059838 ("nvmet-fcloop: Check remoteport port_state before calling done callback")
The above two fixes are just merged to 7.0-rc3, and I will address them by creating new jira issues.
Signed-off-by: Ming Lei [email protected]
Hi, @minlei,
In your backport of commit 858299dc61603 ("block: add scatterlist-less DMA mapping helpers"), I can't parse your Conflicts explanation:
Conflicts: pull block/blk-mq-dma.c change in commit 072cf49380a3 ("PCI/P2PDMA: Simplify bus address mapping API") when this delta change is dropped because this commit isn't added yet
Can you please rewrite that with the correct upstream hash (d4504262f745e), and with a more understandable commit message? Maybe something like the following:
"RHEL backported commit d4504262f745e ("PCI/P2PDMA: Simplify bus address mapping API") out of order. As a result, this patch uses the new calling convention for pci_p2pdma_bus_addr_map()."
You backported commit 027a7a9c07d0d ("drbd: init queue_limits->max_hw_wzeroes_unmap_sectors parameter"), but we don't have the commit that it fixes (0c40d7cb5ef3).
Other patches to consider:
ff20487308f41 ("bio: use memzero_page() in bio_truncate()")
5f33b5226c9d9 ("block: don't autoload drivers on stat").
ffa1e7ada4560 ("block: Make request_queue lockdep splats show up earlier")?
6564862d646e7 ("block: Ensure start sector is aligned for stacking atomic writes")
2729a60bbfb92 ("block: don't silently ignore metadata for sync read/write")
This backport also leaves out the support for block size > page size (original patch posting here: https://lore.kernel.org/all/[email protected]/). I'm ok with that, but I do want to make sure that was intentional.
block: fix memory leak in __blkdev_issue_zero_pages
JIRA: https://issues.redhat.com/browse/RHEL-137852
Signed-off-by: Ming Lei [email protected]