Kolbjørn and Jonáš reported that their 32-bit PowerMacs were crashing
in pata-macio since commit 09fe2bfa ("ata: pata_macio: Fix
max_segment_size with PAGE_SIZE == 64K").

For example:

  kernel BUG at drivers/ata/pata_macio.c:544!
  Oops: Exception in kernel mode, sig: 5 [#1]
  BE PAGE_SIZE=4K MMU=Hash SMP NR_CPUS=2 DEBUG_PAGEALLOC PowerMac
  ...
  NIP pata_macio_qc_prep+0xf4/0x190
  LR  pata_macio_qc_prep+0xfc/0x190
  Call Trace:
    0xc1421660 (unreliable)
    ata_qc_issue+0x14c/0x2d4
    __ata_scsi_queuecmd+0x200/0x53c
    ata_scsi_queuecmd+0x50/0xe0
    scsi_queue_rq+0x788/0xb1c
    __blk_mq_issue_directly+0x58/0xf4
    blk_mq_plug_issue_direct+0x8c/0x1b4
    blk_mq_flush_plug_list.part.0+0x584/0x5e0
    __blk_flush_plug+0xf8/0x194
    __submit_bio+0x1b8/0x2e0
    submit_bio_noacct_nocheck+0x230/0x304
    btrfs_work_helper+0x200/0x338
    process_one_work+0x1a8/0x338
    worker_thread+0x364/0x4c0
    kthread+0x100/0x104
    start_kernel_thread+0x10/0x14

That commit increased max_segment_size to 64KB, with the justification
that the SCSI core was already using that size when PAGE_SIZE == 64KB,
and that there was existing logic to split over-sized requests.

However with a sufficiently large request, the splitting logic causes
each sg to be split into two commands in the DMA table, leading to
overflow of the DMA table, triggering the BUG_ON().

With default settings the bug doesn't trigger, because the request size
is limited by max_sectors_kb == 1280, however max_sectors_kb can be
increased, and apparently some distros do that by default using udev
rules.

Fix the bug for 4KB kernels by reverting to the old max_segment_size.

For 64KB kernels the sg_tablesize needs to be halved, to allow for the
possibility that each sg will be split into two.

Fixes: 09fe2bfa ("ata: pata_macio: Fix max_segment_size with PAGE_SIZE == 64K")
Cc: stable@vger.kernel.org # v6.10+
Reported-by: Kolbjørn Barmen <linux-ppc@kolla.no>
Closes: https://lore.kernel.org/all/62d248bb-e97a-25d2-bcf2-9160c518cae5@kolla.no/Reported-by: Jonáš Vidra <vidra@ufal.mff.cuni.cz>
Closes: https://lore.kernel.org/all/3b6441b8-06e6-45da-9e55-f92f2c86933e@ufal.mff.cuni.cz/Tested-by: Kolbjørn Barmen <linux-ppc@kolla.no>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>