[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <32a7f1ad-e28a-4494-9293-96237c4ed70b@kernel.org>
Date: Tue, 6 May 2025 11:29:08 +0900
From: Damien Le Moal <dlemoal@...nel.org>
To: Steve Siwinski <stevensiwinski@...il.com>, hch@...radead.org
Cc: James.Bottomley@...senpartnership.com, bgrove@...o.com,
linux-kernel@...r.kernel.org, linux-scsi@...r.kernel.org,
martin.petersen@...cle.com, ssiwinski@...o.com, axboe@...nel.dk,
tdoedline@...o.com, linux-block@...r.kernel.org
Subject: Re: [PATCH v2] block, scsi: sd_zbc: Respect bio vector limits for
report zones buffer
On 5/3/25 04:35, Steve Siwinski wrote:
> The report zones buffer size is currently limited by the HBA's
> maximum segment count to ensure the buffer can be mapped. However,
> the block layer further limits the number of iovec entries to
> 1024 when allocating a bio.
>
> To avoid allocation of buffers too large to be mapped, further
> restrict the maximum buffer size to BIO_MAX_INLINE_VECS.
>
> Replace the UIO_MAXIOV symbolic name with the more contextually
> appropriate BIO_MAX_INLINE_VECS.
>
> Signed-off-by: Steve Siwinski <ssiwinski@...o.com>
This needs a "Fixes" tag:
Fixes: b091ac616846 ("sd_zbc: Fix report zones buffer allocation")
Cc: stable@...r.kernel.org
> diff --git a/drivers/scsi/sd_zbc.c b/drivers/scsi/sd_zbc.c
> index 7a447ff600d2..a5364fdc2824 100644
> --- a/drivers/scsi/sd_zbc.c
> +++ b/drivers/scsi/sd_zbc.c
> @@ -180,12 +180,15 @@ static void *sd_zbc_alloc_report_buffer(struct scsi_disk *sdkp,
> * Furthermore, since the report zone command cannot be split, make
> * sure that the allocated buffer can always be mapped by limiting the
> * number of pages allocated to the HBA max segments limit.
> + * Since max segments can be larger than the max inline bio vectors,
> + * further limit the allocated buffer to BIO_MAX_INLINE_VECS.
> */
> nr_zones = min(nr_zones, sdkp->zone_info.nr_zones);
> bufsize = roundup((nr_zones + 1) * 64, SECTOR_SIZE);
> bufsize = min_t(size_t, bufsize,
> queue_max_hw_sectors(q) << SECTOR_SHIFT);
> bufsize = min_t(size_t, bufsize, queue_max_segments(q) << PAGE_SHIFT);
> + bufsize = min_t(size_t, bufsize, BIO_MAX_INLINE_VECS << PAGE_SHIFT);
I would prefer something like:
unsigned int max_segments;
...
max_segments = min(BIO_MAX_INLINE_VECS, queue_max_segments(q));
bufsize = min_t(size_t, bufsize, max_segments << PAGE_SHIFT);
>
> while (bufsize >= SECTOR_SIZE) {
> buf = kvzalloc(bufsize, GFP_KERNEL | __GFP_NORETRY);
> diff --git a/include/linux/bio.h b/include/linux/bio.h
> index cafc7c215de8..7cf9506a6c36 100644
> --- a/include/linux/bio.h
> +++ b/include/linux/bio.h
> @@ -11,6 +11,8 @@
> #include <linux/uio.h>
>
> #define BIO_MAX_VECS 256U
> +/* BIO_MAX_INLINE_VECS must be at most the size of UIO_MAXIOV */
> +#define BIO_MAX_INLINE_VECS 1024
This should be:
#define BIO_MAX_INLINE_VECS UIO_MAXIOV
so that we do not end up with inconsistencies with what user space sees as the
maximum value.
>
> struct queue_limits;
>
--
Damien Le Moal
Western Digital Research
Powered by blists - more mailing lists