Re: [PATCH 1/2] nbd: implement the WRITE_ZEROES command
From: Wouter Verhelst
Date: Tue Aug 06 2024 - 03:50:26 EST
On Mon, Aug 05, 2024 at 07:52:42AM -0500, Eric Blake wrote:
> On Sat, Aug 03, 2024 at 03:04:30PM GMT, Wouter Verhelst wrote:
> > The NBD protocol defines a message for zeroing out a region of an export
> >
> > Add support to the kernel driver for that message.
> >
> > Signed-off-by: Wouter Verhelst <w@xxxxxxx>
> > ---
> > drivers/block/nbd.c | 8 ++++++++
> > include/uapi/linux/nbd.h | 5 ++++-
> > 2 files changed, 12 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/block/nbd.c b/drivers/block/nbd.c
> > index 5b1811b1ba5f..215e7ea9a3c3 100644
> > --- a/drivers/block/nbd.c
> > +++ b/drivers/block/nbd.c
> > @@ -352,6 +352,8 @@ static int __nbd_set_size(struct nbd_device *nbd, loff_t bytesize,
> > }
> > if (nbd->config->flags & NBD_FLAG_ROTATIONAL)
> > lim.features |= BLK_FEAT_ROTATIONAL;
> > + if (nbd->config->flags & NBD_FLAG_SEND_WRITE_ZEROES)
> > + lim.max_write_zeroes_sectors = UINT_MAX;
>
> Is that number accurate, when the kernel has not yet been taught to
> use 64-bit transactions and can therefore only request a 32-bit byte
> length on any one transaction? Would a better limit be
> UINT_MAX/blksize?
Thanks, good catch.
I copied the logic from the handling of the TRIM command (i.e., the
discard logic), which has the same flawed UINT_MAX behavior. I will fix
this in v2 and add a fix for the discard code.
--
w@uter.{be,co.za}
wouter@{grep.be,fosdem.org,debian.org}
I will have a Tin-Actinium-Potassium mixture, thanks.