The comment is true for all stable versions.
This patch is part of a larger series that enables ESE volumes.
I think it should not go alone as other patches like
5e2b17e712cf s390/dasd: Add dynamic formatting support for ESE volumes
are needed to actually work with ESE volumes.
So I suggest to drop this patch.
Jan, Stefan, do you agree?
On 19.07.19 05:56, Sasha Levin wrote:
> From: Jan Höppner <hoepp...@linux.ibm.com>
>
> [ Upstream commit ce6915f5343f5f2a2a937b683d8ffbf12dab3ad4 ]
>
> The disk layout and volume information of a DASD reside in the first two
> tracks of cylinder 0. When a DASD is set online, currently the first
> three tracks are read and analysed to confirm an expected layout.
>
> For CDL (Compatible Disk Layout) only count area data of the first track
> is evaluated and checked against expected key and data lengths. For LDL
> (Linux Disk Layout) the first and third track is evaluated. However,
> an LDL formatted volume is expected to be in the same format across all
> tracks. Checking the third track therefore doesn't have any more value
> than checking any other track at random.
>
> Now, an Extent Space Efficient (ESE) DASD is initialised by only
> formatting the first two tracks, as those tracks always contain all
> information necessarry.
>
> Checking the third track on an ESE volume will therefore most likely
> fail with a record not found error, as the third track will be empty.
> This in turn leads to the device being recognised with a volume size of
> 0. Attempts to write volume information on the first two tracks then
> fail with "no space left on device" errors.
>
> Initialising the first three tracks for an ESE volume is not a viable
> solution, because the third track is already a regular track and could
> contain user data. With that there is potential for data corruption.
>
> Instead, always only analyse the first two tracks, as it is sufficiant
> for both CDL and LDL, and allow ESE volumes to be recognised as well.
>
> Signed-off-by: Jan Höppner <hoepp...@linux.ibm.com>
> Reviewed-by: Stefan Haberland <s...@linux.ibm.com>
> Signed-off-by: Vasily Gorbik <g...@linux.ibm.com>
> Signed-off-by: Sasha Levin <sas...@kernel.org>
> ---
> drivers/s390/block/dasd_eckd.c | 12 ++++++------
> 1 file changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/drivers/s390/block/dasd_eckd.c b/drivers/s390/block/dasd_eckd.c
> index c09039eea707..c7aec1b44b7c 100644
> --- a/drivers/s390/block/dasd_eckd.c
> +++ b/drivers/s390/block/dasd_eckd.c
> @@ -157,7 +157,7 @@ static const int sizes_trk0[] = { 28, 148, 84 };
> #define LABEL_SIZE 140
>
> /* head and record addresses of count_area read in analysis ccw */
> -static const int count_area_head[] = { 0, 0, 0, 0, 2 };
> +static const int count_area_head[] = { 0, 0, 0, 0, 1 };
> static const int count_area_rec[] = { 1, 2, 3, 4, 1 };
>
> static inline unsigned int
> @@ -1823,8 +1823,8 @@ dasd_eckd_analysis_ccw(struct dasd_device *device)
> if (IS_ERR(cqr))
> return cqr;
> ccw = cqr->cpaddr;
> - /* Define extent for the first 3 tracks. */
> - define_extent(ccw++, cqr->data, 0, 2,
> + /* Define extent for the first 2 tracks. */
> + define_extent(ccw++, cqr->data, 0, 1,
> DASD_ECKD_CCW_READ_COUNT, device, 0);
> LO_data = cqr->data + sizeof(struct DE_eckd_data);
> /* Locate record for the first 4 records on track 0. */
> @@ -1843,9 +1843,9 @@ dasd_eckd_analysis_ccw(struct dasd_device *device)
> count_data++;
> }
>
> - /* Locate record for the first record on track 2. */
> + /* Locate record for the first record on track 1. */
> ccw[-1].flags |= CCW_FLAG_CC;
> - locate_record(ccw++, LO_data++, 2, 0, 1,
> + locate_record(ccw++, LO_data++, 1, 0, 1,
> DASD_ECKD_CCW_READ_COUNT, device, 0);
> /* Read count ccw. */
> ccw[-1].flags |= CCW_FLAG_CC;
> @@ -1967,7 +1967,7 @@ static int dasd_eckd_end_analysis(struct dasd_block
> *block)
> }
> }
> if (i == 3)
> - count_area = &private->count_area[4];
> + count_area = &private->count_area[3];
>
> if (private->uses_cdl == 0) {
> for (i = 0; i < 5; i++) {
>
