[PATCHv2] nvmet: lift blocksize restriction to 64k

[PATCHv2] nvmet: lift blocksize restriction to 64k

[Hannes Reinecke]

With commit 47dd67532303 ("block/bdev: lift block size restrictions to 64k")
we can now support up to 64k block sizes, so lift the restriction on 4k for
file-backed namespaces.

Reported-by: Abdulla Nasirli <nasirabd@fit.cvut.cz>
Fixes: 47dd67532303 ("block/bdev: lift block size restrictions to 64k")
Signed-off-by: Hannes Reinecke <hare@kernel.org>
---
 drivers/nvme/target/io-cmd-file.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/drivers/nvme/target/io-cmd-file.c b/drivers/nvme/target/io-cmd-file.c
index 2d068439b129..fdc15c59632f 100644
--- a/drivers/nvme/target/io-cmd-file.c
+++ b/drivers/nvme/target/io-cmd-file.c
@@ -50,11 +50,10 @@ int nvmet_file_ns_enable(struct nvmet_ns *ns)
 	nvmet_file_ns_revalidate(ns);
 
 	/*
-	 * i_blkbits can be greater than the universally accepted upper bound,
-	 * so make sure we export a sane namespace lba_shift.
+	 * We can support blocksizes up to 64k.
 	 */
 	ns->blksize_shift = min_t(u8,
-			file_inode(ns->file)->i_blkbits, 12);
+			file_inode(ns->file)->i_blkbits, ilog2(SZ_64K));
 
 	ns->bvec_pool = mempool_create(NVMET_MIN_MPOOL_OBJ, mempool_alloc_slab,
 			mempool_free_slab, nvmet_bvec_cache);
-- 
2.43.0

Re: [PATCHv2] nvmet: lift blocksize restriction to 64k

[Christoph Hellwig]

On Fri, Feb 20, 2026 at 04:12:59PM +0100, Hannes Reinecke wrote:
> With commit 47dd67532303 ("block/bdev: lift block size restrictions to 64k")
> we can now support up to 64k block sizes, so lift the restriction on 4k for
> file-backed namespaces.

Still not sure this is a great idea by default.  Yes, modern Linux supports
64k sector sizes, but even then just for a few file systems.  And this
changes existing setups, which will now break when upgrading the kernel.

This needs an opt-in.

Re: [PATCHv2] nvmet: lift blocksize restriction to 64k

[Hannes Reinecke]

On 2/20/26 16:32, Christoph Hellwig wrote:
> On Fri, Feb 20, 2026 at 04:12:59PM +0100, Hannes Reinecke wrote:
>> With commit 47dd67532303 ("block/bdev: lift block size restrictions to 64k")
>> we can now support up to 64k block sizes, so lift the restriction on 4k for
>> file-backed namespaces.
> 
> Still not sure this is a great idea by default.  Yes, modern Linux supports
> 64k sector sizes, but even then just for a few file systems.  And this
> changes existing setups, which will now break when upgrading the kernel.
> 
How so? There cannot be any existing nvmet setup with an LBS backing 
device, as the device will be rejected when trying to setup the target.
Which breakage do you see here?

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                  Kernel Storage Architect
hare@suse.de                                +49 911 74053 688
SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg
HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich

Re: [PATCHv2] nvmet: lift blocksize restriction to 64k

[Christoph Hellwig]

On Fri, Feb 20, 2026 at 05:42:25PM +0100, Hannes Reinecke wrote:
> On 2/20/26 16:32, Christoph Hellwig wrote:
>> On Fri, Feb 20, 2026 at 04:12:59PM +0100, Hannes Reinecke wrote:
>>> With commit 47dd67532303 ("block/bdev: lift block size restrictions to 64k")
>>> we can now support up to 64k block sizes, so lift the restriction on 4k for
>>> file-backed namespaces.
>>
>> Still not sure this is a great idea by default.  Yes, modern Linux supports
>> 64k sector sizes, but even then just for a few file systems.  And this
>> changes existing setups, which will now break when upgrading the kernel.
>>
> How so? There cannot be any existing nvmet setup with an LBS backing 
> device, as the device will be rejected when trying to setup the target.
> Which breakage do you see here?

There can be a setup with a large i_blkbits that we limit to 4k
export LBA size right now, and with your patch it will be exposed
as something larger.

Re: [PATCHv2] nvmet: lift blocksize restriction to 64k

[Hannes Reinecke]

On 2/20/26 17:45, Christoph Hellwig wrote:
> On Fri, Feb 20, 2026 at 05:42:25PM +0100, Hannes Reinecke wrote:
>> On 2/20/26 16:32, Christoph Hellwig wrote:
>>> On Fri, Feb 20, 2026 at 04:12:59PM +0100, Hannes Reinecke wrote:
>>>> With commit 47dd67532303 ("block/bdev: lift block size restrictions to 64k")
>>>> we can now support up to 64k block sizes, so lift the restriction on 4k for
>>>> file-backed namespaces.
>>>
>>> Still not sure this is a great idea by default.  Yes, modern Linux supports
>>> 64k sector sizes, but even then just for a few file systems.  And this
>>> changes existing setups, which will now break when upgrading the kernel.
>>>
>> How so? There cannot be any existing nvmet setup with an LBS backing
>> device, as the device will be rejected when trying to setup the target.
>> Which breakage do you see here?
> 
> There can be a setup with a large i_blkbits that we limit to 4k
> export LBA size right now, and with your patch it will be exposed
> as something larger.
> 
True.
Hmm. Okay, I'll work on something.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                  Kernel Storage Architect
hare@suse.de                                +49 911 74053 688
SUSE Software Solutions GmbH, Frankenstr. 146, 90461 Nürnberg
HRB 36809 (AG Nürnberg), GF: I. Totev, A. McDonald, W. Knoblich