Couple of fixes detected with blktests
Subsystems (2)
Sagi Grimberg
maintainerv4 cover Couple of fixes detected with blktests
nvme test 009 discovered a few issues with and without multipathing.
1. if a namespace has a larger block size than PAGE_SIZE we either panic on a divide by zero with multipathing, or we see I/O errors without it.
- fix nvmet to always export block size that is leq PAGE_SIZE
- fix the core to set 0 capacity on such a namespace
2. if a namespace allocation fails after nvme_init_ns_head we leak subsystems due to a missing ns_head ref put
- fix it by putting the ns_head reference on nvme_alloc_ns error flow
Changes from v3:
- don't have the host fail if ns allocation lba_shift exceeds PAGE_SHIFT, instead set 0 capacity
Changes from v2:
- add comment to nvmet file ns lba_shift clamping
- add host side protection against such a namespace
Sagi Grimberg (3):
nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
nvme: put ns_head ref if namespace fails allocation
nvmet-file: clamp-down file namespace lba_shift
drivers/nvme/host/core.c | 8 +++++++- drivers/nvme/target/io-cmd-file.c | 7 ++++++- 2 files changed, 13 insertions(+), 2 deletions(-)
Signature · 1 line
2.17.1
Sagi Grimberg
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
If our target exposed a namespace with a block size that is greater than PAGE_SIZE, set 0 capacity on the namespace as we do not support it.
This issue encountered when the nvmet namespace was backed by a tempfile.
--- drivers/nvme/host/core.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/nvme/host/core.c +6 -1
@@ -1610,6 +1610,10 @@ static void nvme_update_disk_info(struct gendisk *disk,sector_t capacity =le64_to_cpup (&id->nsze) << (ns->lba_shift -9 );unsigned short bs =1 << ns->lba_shift; +if (ns->lba_shift >PAGE_SHIFT ) { +/* unsupported block size, set capacity to 0 later */ + bs = (1 <<9 ); + }blk_mq_freeze_queue (disk->queue);blk_integrity_unregister (disk);
@@ -1621,7 +1625,8 @@ static void nvme_update_disk_info(struct gendisk *disk,if (ns->ms && !ns->ext && (ns->ctrl->ops->flags &NVME_F_METADATA_SUPPORTED ))nvme_init_integrity (disk, ns->ms, ns->pi_type); -if (ns->ms && !nvme_ns_has_pi (ns) && !blk_get_integrity (disk)) +if ((ns->ms && !nvme_ns_has_pi (ns) && !blk_get_integrity (disk)) || + ns->lba_shift >PAGE_SHIFT ) capacity =0 ;set_capacity (disk, capacity);
Signature · 1 line
2.17.1
PATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
Quoted 6 lines · Sagi Grimberg
If our target exposed a namespace with a block size that is greater than
PAGE_SIZE, set 0 capacity on the namespace as we do not support it.This issue encountered when the nvmet namespace was backed by a tempfile.
Signed-off-by Sagi Grimberg <sagi at grimberg.me>
I responded to the v3 with the review by mistake. :)
Christoph Hellwig
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
In reply to Sagi GrimbergQuoted 26 lines · Sagi Grimberg
If our target exposed a namespace with a block size that is greater than
PAGE_SIZE, set 0 capacity on the namespace as we do not support it.This issue encountered when the nvmet namespace was backed by a tempfile.
Signed-off-by Sagi Grimberg <sagi at grimberg.me>--- drivers/nvme/host/core.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)drivers/nvme/host/core.c +4
-0@@ -1610,6 +1610,10 @@ static void nvme_update_disk_info(struct gendisk *disk,sector_t capacity =le64_to_cpup (&id->nsze) << (ns->lba_shift -9 );unsigned short bs =1 << ns->lba_shift; +if (ns->lba_shift >PAGE_SHIFT ) { +/* unsupported block size, set capacity to 0 later */ + bs = (1 <<9 ); + }blk_mq_freeze_queue (disk->queue);blk_integrity_unregister (disk);
drivers/nvme/host/core.c static void nvme_update_disk_info(struct gendisk *disk, · line 1630
Quoted 7 lines
drivers/nvme/host/core.c +2
-1@@ -1621,7 +1625,8 @@ static void nvme_update_disk_info(struct gendisk *disk,if (ns->ms && !ns->ext && (ns->ctrl->ops->flags &NVME_F_METADATA_SUPPORTED ))nvme_init_integrity (disk, ns->ms, ns->pi_type); -if (ns->ms && !nvme_ns_has_pi (ns) && !blk_get_integrity (disk)) +if ((ns->ms && !nvme_ns_has_pi (ns) && !blk_get_integrity (disk)) || + ns->lba_shift >PAGE_SHIFT ) capacity =0 ;
I like the idea behind this, but it looks rather convoluted. I think for the unusable namespace case we should warn and have a common label that just sets the capacity, not touching anything else.
Does something like this work for you?
---
Quoted 1 line
From
ec4ce1708eb4d4f536b93f09dab952067f58ba4bMon Sep 17 00:00:00 2001
From: Christoph Hellwig <[email protected]>
Date: Tue, 12 Mar 2019 15:31:09 +0100
Subject: nvme: ignore namespaces with an unusable block size
And also refactor the handling for unusable namespaces to use a common goto label, and print a warning for invalid metadata formats as well while we are at it.
--- drivers/nvme/host/core.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-)
drivers/nvme/host/core.c +15 -2
@@ -1599,6 +1599,12 @@ static void nvme_update_disk_info(struct gendisk *disk,blk_mq_freeze_queue (disk->queue);blk_integrity_unregister (disk); +if (bs >PAGE_SIZE ) { +dev_warn (ns->ctrl->device, +"unsupported LBA size %u, ignoring namespace\n" , bs); + goto ignore; + } +blk_queue_logical_block_size (disk->queue, bs);blk_queue_physical_block_size (disk->queue, bs);blk_queue_io_min (disk->queue, bs);
@@ -1606,8 +1612,11 @@ static void nvme_update_disk_info(struct gendisk *disk,if (ns->ms && !ns->ext && (ns->ctrl->ops->flags &NVME_F_METADATA_SUPPORTED ))nvme_init_integrity (disk, ns->ms, ns->pi_type); -if (ns->ms && !nvme_ns_has_pi (ns) && !blk_get_integrity (disk)) - capacity =0 ; +if (ns->ms && !nvme_ns_has_pi (ns) && !blk_get_integrity (disk)) { +dev_warn (ns->ctrl->device, +"unsupported metadata format, ignoring namespace\n" ); + goto ignore; + }set_capacity (disk, capacity);nvme_ns_config_oncs (ns);
@@ -1618,6 +1627,10 @@ static void nvme_update_disk_info(struct gendisk *disk,set_disk_ro (disk, false);blk_mq_unfreeze_queue (disk->queue); +return ; +ignore: +blk_mq_unfreeze_queue (disk->queue); +set_capacity (disk,0 ); }static void __nvme_revalidate_disk (struct gendisk *disk,struct nvme_id_ns *id)
Signature · 1 line
2.20.1
Christoph Hellwig
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
And while we are at it - Dave Chinner has been actively working on blocksize > PAGE_SIZE support for XFS. So we might be able to use "too large" sectors sizes very soon.
Sagi Grimberg
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
In reply to Christoph HellwigQuoted 6 lines
I like the idea behind this, but it looks rather convoluted. I think for the unusable namespace case we should warn and have a common label that just sets the capacity, not touching anything else.
Does something like this work for you?
No, this is what I had done originally, but we need to always have the queue set to a decent block size, otherwise blk_queue_stack_limits() panics on div by 0..
Christoph Hellwig
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
Quoted 10 lines · Sagi Grimberg
Quoted 6 lines
I like the idea behind this, but it looks rather convoluted. I think for the unusable namespace case we should warn and have a common label that just sets the capacity, not touching anything else.
Does something like this work for you?
No, this is what I had done originally, but we need to always have the queue set to a decent block size, otherwise
blk_queue_stack_limits()panics on div by 0..
I actually tested it by manually hacking a 8k block size into nvmet and and it works just fine for me. Where do you see a division by zero with this patch exactly?
Keith Busch
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
Quoted 8 lines · Christoph Hellwig
Quoted 7 lines · Sagi Grimberg
Quoted 6 lines
I like the idea behind this, but it looks rather convoluted. I think for the unusable namespace case we should warn and have a common label that just sets the capacity, not touching anything else.
Does something like this work for you?
Quoted 7 lines
Quoted 3 lines
No, this is what I had done originally, but we need to always have the queue set to a decent block size, otherwise
blk_queue_stack_limits()panics on div by 0..I actually tested it by manually hacking a 8k block size into nvmet and and it works just fine for me. Where do you see a division by zero with this patch exactly?
I'm not sure abuot a divide-by-zero, but I just hacked up qemu to report an 8k block size and get this on boot. Happens because alloc_page_buffers won't allocate a buffer_head when block size is greater than a page size:
BUG: unable to handle kernel NULL pointer dereference at 0000000000000008 #PF error: [normal kernel read fault] PGD 0 P4D 0 Oops: 0000 [#1] SMP CPU: 3 PID: 391 Comm: kworker/u18:1 Not tainted 5.0.0+ #42 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.0-0-ga698c8995f-prebuilt.qemu.org 04/01/2014 Workqueue: nvme-wq nvme_scan_work [nvme_core] RIP: 0010:create_empty_buffers+0x24/0x100 Code: eb cb 0f 1f 40 00 0f 1f 44 00 00 41 54 55 49 89 d4 53 ba 01 00 00 00 48 89 fb e8 87 fe ff ff 48 89 c5 48 89 c2 eb 03 48 89 ca <48> 8b 4a 08 4c 09 22 48 85 c9 75 f1 48 89 6a 08 48 8b 43 18 48 8d RSP: 0018:ffffbd9ec05cf880 EFLAGS: 00010286 RAX: 0000000000000000 RBX: ffffe0fec03a38c0 RCX: ffff99f4c751d000 RDX: 0000000000000000 RSI: ffff99f4c751d000 RDI: ffffe0fec03a38c0 RBP: 0000000000000000 R08: dead0000000000ff R09: 0000000000000003 R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000 R13: 0000000000000000 R14: 0000000000000200 R15: ffffe0fec03a38c0 FS: 0000000000000000(0000) GS:ffff99f4cfd00000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 0000000000000008 CR3: 000000000eb28000 CR4: 00000000000006e0 DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 Call Trace: create_page_buffers+0x4d/0x60 block_read_full_page+0x47/0x310 ? __add_to_page_cache_locked+0x288/0x330 ? check_disk_change+0x60/0x60 ? count_shadow_nodes+0x130/0x130 do_read_cache_page+0x31c/0x6b0 ? blkdev_writepages+0x10/0x10 read_dev_sector+0x28/0xc0 read_lba+0x126/0x210 ? kmem_cache_alloc_trace+0x19b/0x1b0 efi_partition+0x137/0x780 ? vsnprintf+0x2ae/0x4a0 ? vsnprintf+0xec/0x4a0 ? snprintf+0x45/0x70 ? is_gpt_valid.part.6+0x400/0x400 ? check_partition+0x137/0x240 check_partition+0x137/0x240 rescan_partitions+0xab/0x350 __blkdev_get+0x342/0x560 ? inode_insert5+0x11f/0x1e0 blkdev_get+0x11f/0x310 ? unlock_new_inode+0x44/0x60 ? bdget+0xff/0x110 __device_add_disk+0x426/0x470 nvme_validate_ns+0x35e/0x7c0 [nvme_core] ? nvme_identify_ctrl.isra.56+0x7e/0xc0 [nvme_core] ? update_load_avg+0x89/0x550 nvme_scan_work+0xe5/0x370 [nvme_core] ? __synchronize_srcu.part.18+0x91/0xc0 ? try_to_wake_up+0x55/0x430 process_one_work+0x1e9/0x3e0 worker_thread+0x21a/0x3d0 ? process_one_work+0x3e0/0x3e0 kthread+0x111/0x130 ? kthread_park+0x90/0x90 ret_from_fork+0x1f/0x30 Modules linked in: nvme nvme_core serio_raw CR2: 0000000000000008 ---[ end trace b38bdf1b424f36e9 ]---
Christoph Hellwig
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
Quoted 13 lines · Keith Busch
Quoted 12 lines · Christoph Hellwig
Quoted 10 lines · Sagi Grimberg
Quoted 6 lines
I like the idea behind this, but it looks rather convoluted. I think for the unusable namespace case we should warn and have a common label that just sets the capacity, not touching anything else.
Does something like this work for you?
No, this is what I had done originally, but we need to always have the queue set to a decent block size, otherwise
blk_queue_stack_limits()panics on div by 0..
Quoted 7 lines
Quoted 3 lines
I actually tested it by manually hacking a 8k block size into nvmet and and it works just fine for me. Where do you see a division by zero with this patch exactly?
I'm not sure abuot a divide-by-zero, but I just hacked up qemu to report an 8k block size and get this on boot. Happens because
alloc_page_bufferswon't allocate abuffer_headwhen block size is greater than a page size:
But that is without either the patch from Sagi or the one from me, right?
Keith Busch
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
Quoted 7 lines · Christoph Hellwig
Quoted 3 lines · Keith Busch
I'm not sure abuot a divide-by-zero, but I just hacked up qemu to report an 8k block size and get this on boot. Happens because
alloc_page_bufferswon't allocate abuffer_headwhen block size is greater than a page size:But that is without either the patch from Sagi or the one from me, right?
Sorry, I misunderstood to think you were saying only that Sagi's patch wasn't necessary. Yes, either patch works.
Sagi Grimberg
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
In reply to Christoph HellwigQuoted 7 lines
Quoted 7 lines
Quoted 6 lines
I like the idea behind this, but it looks rather convoluted. I think for the unusable namespace case we should warn and have a common label that just sets the capacity, not touching anything else.
Does something like this work for you?
Quoted 7 lines
Quoted 3 lines
No, this is what I had done originally, but we need to always have the queue set to a decent block size, otherwise
blk_queue_stack_limits()panics on div by 0..I actually tested it by manually hacking a 8k block size into nvmet and and it works just fine for me. Where do you see a division by zero with this patch exactly?
You need your block size to exceed u16 bs.
Then run blktest/tests/nvme/009 with multipath on.
Christoph Hellwig
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
Quoted 11 lines · Sagi Grimberg
Quoted 11 lines
Quoted 10 lines
Quoted 6 lines
I like the idea behind this, but it looks rather convoluted. I think for the unusable namespace case we should warn and have a common label that just sets the capacity, not touching anything else.
Does something like this work for you?
No, this is what I had done originally, but we need to always have the queue set to a decent block size, otherwise
blk_queue_stack_limits()panics on div by 0..
Quoted 5 lines
Quoted 3 lines
I actually tested it by manually hacking a 8k block size into nvmet and and it works just fine for me. Where do you see a division by zero with this patch exactly?
You need your block size to exceed u16 bs.
I'll need to look into actually testing something that large, but the whole point of my patch is that we never even use the bs value for anything. So I'm not sure why that would matter.
Sagi Grimberg
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
Quoted 8 lines
Quoted 6 lines
Quoted 4 lines
I actually tested it by manually hacking a 8k block size into nvmet and and it works just fine for me. Where do you see a division by zero with this patch exactly?
You need your block size to exceed u16 bs.
I'll need to look into actually testing something that large,
just run blktests nvme/009 test with multipath, nothing fancy (I used ext4).
From a quick look I think that ext4 tempfiles are coming from
the transaction subsystem so maybe that's why the inode i_blkbits is 20? Not sure at all though...
Anyways, its very easy to reproduce.
Quoted 2 lines
but the whole point of my patch is that we never even use the bs value for anything. So I'm not sure why that would matter.
But than with multipath, blk_queue_stack_limits assume that we have a valid bs in the lower device queue.
Sagi Grimberg
maintainerPATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
Quoted 15 lines
Quoted 7 lines
Quoted 5 lines
Quoted 3 lines
I actually tested it by manually hacking a 8k block size into nvmet and and it works just fine for me.? Where do you see a division by zero with this patch exactly?
You need your block size to exceed u16 bs.
I'll need to look into actually testing something that large,
just run blktests nvme/009 test with multipath, nothing fancy (I used ext4).
From a quick look I think that ext4 tempfiles are coming from
the transaction subsystem so maybe that's why the inodei_blkbitsis 20? Not sure at all though...
Quoted 7 lines
Anyways, its very easy to reproduce.
Quoted 2 lines
but the whole point of my patch is that we never even use the bs value for anything. So I'm not sure why that would matter.
But than with multipath,
blk_queue_stack_limitsassume that we have a valid bs in the lower device queue.
Reviving this,
This is still failing without this patch.
Reproducer:
- use ext4 filesystem
- set nvme.multipath=1
- run blktests: tests/nvme/009
bang...
Christoph,
The reason why bs is set to 512 in nvme_update_disk_info() is because blk_queue_stack_limits assumes that q->limits.logical_block_size of the underlying device is not zero (automatically divides by it).
PATCH v4 1/3 nvme: set 0 capacity if namespace block size exceeds PAGE_SIZE
In reply to Sagi GrimbergThanks,
applied to nvme-5.2.
Sagi Grimberg
maintainerPATCH v4 2/3 nvme: put ns_head ref if namespace fails allocation
In reply to Sagi GrimbergIn case nvme_alloc_ns fails after we initialize ns_head but before we add the ns to the controller namespaces list we need to explicitly put the ns_head reference because when we teardown the controller we won't find it, causing us to leak a dangling subsystem eventually.
--- drivers/nvme/host/core.c | 1 + 1 file changed, 1 insertion(+)
drivers/nvme/host/core.c +1 -0
@@ -3314,6 +3314,7 @@ static void nvme_alloc_ns(struct nvme_ctrl *ctrl, unsigned nsid)mutex_lock (&ctrl->subsys->lock);list_del_rcu (&ns->siblings);mutex_unlock (&ctrl->subsys->lock); +nvme_put_ns_head (ns->head); out_free_id:kfree (id); out_free_queue:
Signature · 1 line
2.17.1
PATCH v4 2/3 nvme: put ns_head ref if namespace fails allocation
Thanks,
applied to nvme-5.1.
Sagi Grimberg
maintainerPATCH v4 3/3 nvmet-file: clamp-down file namespace lba_shift
In reply to Sagi GrimbergWhen the backing file is a tempfile for example, the inode i_blkbits can be 1M in size which causes problems for hosts to support as the disk block size.
--- drivers/nvme/target/io-cmd-file.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)
drivers/nvme/target/io-cmd-file.c +6 -1
@@ -49,7 +49,12 @@ int nvmet_file_ns_enable(struct nvmet_ns *ns) goto err; ns->size = stat.size; - ns->blksize_shift =file_inode (ns->file)->i_blkbits; +/* +* inode i_blkbits can be greater than the universally accepted upper +* bound so make sure we export we export a sane namespace lba_shift. +*/ + ns->blksize_shift =min_t (u8, +file_inode (ns->file)->i_blkbits,PAGE_SHIFT ); ns->bvec_cache = kmem_cache_create("nvmet-bvec", NVMET_MAX_MPOOL_BVEC * sizeof(struct bio_vec),
Signature · 1 line
2.17.1
Christoph Hellwig
maintainerPATCH v4 3/3 nvmet-file: clamp-down file namespace lba_shift
Quoted 14 lines · Sagi Grimberg
When the backing file is a tempfile for example, the inode
i_blkbitscan be 1M in size which causes problems for hosts to support as the disk block size.Signed-off-by Sagi Grimberg <sagi at grimberg.me>--- drivers/nvme/target/io-cmd-file.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)drivers/nvme/target/io-cmd-file.c +0
-0
drivers/nvme/target/io-cmd-file.c int nvmet_file_ns_enable(struct nvmet_ns *ns) · line 57
Quoted 10 lines
drivers/nvme/target/io-cmd-file.c +6
-1@@ -49,7 +49,12 @@ int nvmet_file_ns_enable(struct nvmet_ns *ns) goto err; ns->size = stat.size; - ns->blksize_shift =file_inode (ns->file)->i_blkbits; +/* +* inode i_blkbits can be greater than the universally accepted upper +* bound so make sure we export we export a sane namespace lba_shift. +*/ + ns->blksize_shift =min_t (u8, +file_inode (ns->file)->i_blkbits,PAGE_SHIFT );
I still wonder why we clamp this down to PAGE SIZE rather than 4k.
Everything larger than 4k has a huge chance of not working with most remote hosts, even if our target system has a larger page size.
Sagi Grimberg
maintainerPATCH v4 3/3 nvmet-file: clamp-down file namespace lba_shift
Quoted 28 lines
Quoted 15 lines
When the backing file is a tempfile for example, the inode
i_blkbitscan be 1M in size which causes problems for hosts to support as the disk block size.Signed-off-by Sagi Grimberg <sagi at grimberg.me>--- drivers/nvme/target/io-cmd-file.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-)drivers/nvme/target/io-cmd-file.c +0
-0drivers/nvme/target/io-cmd-file.c int nvmet_file_ns_enable(struct nvmet_ns *ns) · line 57
Quoted 10 lines
drivers/nvme/target/io-cmd-file.c +6
-1@@ -49,7 +49,12 @@ int nvmet_file_ns_enable(struct nvmet_ns *ns) goto err; ns->size = stat.size; - ns->blksize_shift =file_inode (ns->file)->i_blkbits; +/* +* inode i_blkbits can be greater than the universally accepted upper +* bound so make sure we export we export a sane namespace lba_shift. +*/ + ns->blksize_shift =min_t (u8, +file_inode (ns->file)->i_blkbits,PAGE_SHIFT );I still wonder why we clamp this down to PAGE SIZE rather than 4k.
Quoted 2 lines
Everything larger than 4k has a huge chance of not working with most remote hosts, even if our target system has a larger page size.
loop does the same thing, didn't ever hear complaints on that..
Christoph Hellwig
maintainerPATCH v4 3/3 nvmet-file: clamp-down file namespace lba_shift
Quoted 6 lines · Sagi Grimberg
Quoted 4 lines
I still wonder why we clamp this down to PAGE SIZE rather than 4k.
Everything larger than 4k has a huge chance of not working with most remote hosts, even if our target system has a larger page size.
loop does the same thing, didn't ever hear complaints on that..
loop is always used on the same system exporting things..
Sagi Grimberg
maintainerPATCH v4 3/3 nvmet-file: clamp-down file namespace lba_shift
Quoted 9 lines
Quoted 7 lines
Quoted 5 lines
I still wonder why we clamp this down to PAGE SIZE rather than 4k.
Everything larger than 4k has a huge chance of not working with most remote hosts, even if our target system has a larger page size.
loop does the same thing, didn't ever hear complaints on that..
loop is always used on the same system exporting things..
Yea, I see what you're saying... I can change it to 4k.