Switch from the legacy PCI DMA API to the generic DMA API.
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Reviewed-by: NJohannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

tscam(), atp870_init(), atp880_init() and atp885_init() are never
called in atomic context.
They call mdelay() to busily wait, which is not necessary.
mdelay() can be replaced with msleep().

This is found by a static analysis tool named DCNS written by myself.
Signed-off-by: NJia-Ju Bai <baijiaju1990@gmail.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

On 64 bit CPUs there is a memory corruption bug on probe().  It should
be a u32 pointer instead of an unsigned long pointer or we write past
the end of the setupdata[] array.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: NHannes Reinecke <hare@suse.com>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Move 870-specific init code to a separate function atp870_init()
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Move 885-specific init code to a separate function atp885_init()
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Move 880-specific init code to a separate function atp880_init()
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Call _init_tables before chip-specific initialization. This avoids code
duplication and fixes a bug(?) in 880 init where the values read from flash
into atpdev->sp are then overwritten by calling init_tables.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

scam_on is used only during probe, no need to keep it later.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Allocate IRQ later during probe to avoid code duplication and also
remove the need for weird locking in _probe.
(It was probably there to prevent race with the IRQ handler?)
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Use pci_request_regions and do it before accessing the I/O ports.
Also add missing pci_disable_device() call to atp870u_remove().
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Introduce chip type inline functions to simplify code, allowing to delete
dev_id from struct atp_unit.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Move shpnt common code to the top, remove base_io, use pci_resource_len.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

chip_ver is used for wide chip detection only. Remove it and use a local
variable instead (for 870; 880 and 885 are always wide).
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Move scsi_host_alloc() to the top of _probe() to remove code duplication,
*p and unneeded atpdev (de)allocation and copying.  While at it, fix the
error paths to return real error codes and also add missing
pci_disble_device() call.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

tscam_885() is empty (except a delay) so remove it.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The code for setting host adapter ID is the same for all chips.
Move it to a common function.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

The ioport region is 0x20 bytes long so accessing 0x3a register using
writeb_io is incorrect. Use writeb_base instead.
There's no change in behavior as 870 chips have ioport = baseport.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

tscam() is using port 0x80 access for delays but that's x86-only.
Use udelay(2) instead.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Now that all the is* functions except is885() are gone, rename is885() to
atp_is() to avoid confusion. Don't know what "is" means, though...
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Now that is885() supports everything from is870() and the rest of the code
is almost identical, remove is870() and use is885() instead.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Move few remaining 870-specific code lines out of is870()
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Add remaining 870 support to is885():
 - different synw, no synuw
 - synu[4] = 0x0c
 - atp_writeb_io(dev, c, 0x04, 0x00); instead of
   atp_writeb_io(dev, c, 0x14, 0x00); (isn't that a bug?)
 - atp_writeb_io(dev, c, 0x14, 0xff); instead of
   atp_writeb_io(dev, c, 0x14, 0x06);
 - different mbuf[3] and mbuf[4] checks
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Don't check chip_ver in is870() but add wide_chip parameter for that.
Then add the non-wide support to is885().
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Now that is880() and is885() are almost identical (except for some cpu_relax()
calls and debug printks), remove is880() and use is885() instead.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Move few chip-specifis lines out of is880() and is885() so they become
almost identical.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Add channel parameter to is870() and is880() functions to simplify comparing
them with is885().
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Unify code formatting in is870(), is880() and is885() functions to simplify
comparing them.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Subtract 0x40 to use _io access wrappers.  Now it's obvious that is870()
and is880() are very similar.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Introduce *_read? and *_write? wrappers to improve code readability.
Also make sure that baseport is always initialized, not only for ATP880.
Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>

Signed-off-by: NOndrej Zary <linux@rainbow-software.org>
Reviewed-by: NHannes Reinicke <hare@suse.de>
Acked-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NMartin K. Petersen <martin.petersen@oracle.com>