Skip to contentSkip to conversation

Q40 IDE fixes

PATCH Superseded Michael Schmitz 2 participants 9 messages Aug 18, 2023 – Aug 22, 2023

Subsystems (2)

v3 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v3 cover Q40 IDE fixes

Comments by Finn and Geert have been addressed. Tested-by tag added for both patches.

Patch 3 would have caused user space visible changes, so has been dropped after changes to patch 2 make it handle negative pdev->id correctly.

Cheers,

Michael

#1 lore raw

PATCH v3 1/2 ata: pata_falcon: fix IO base selection for Q40

With commit 44b1fbc0f5f3 ("m68k/q40: Replace q40ide driver with pata_falcon and falconide"), the Q40 IDE driver was replaced by pata_falcon.c.

Both IO and memory resources were defined for the Q40 IDE platform device, but definition of the IDE register addresses was modeled after the Falcon case, both in use of the memory resources and in including register scale and byte vs. word offset in the address.

This was correct for the Falcon case, which does not apply any address translation to the register addresses. In the Q40 case, all of device base address, byte access offset and register scaling is included in the platform specific ISA access translation (in asm/mm_io.h).

As a consequence, such address translation gets applied twice, and register addresses are mangled.

Use the device base address from the platform IO resource, and use standard register offsets from that base in order to calculate register addresses (the IO address translation will then apply the correct ISA window base and scaling).

Encode PIO_OFFSET into IO port addresses for all registers except the data transfer register. Encode the MMIO offset there (pata_falcon_data_xfer() directly uses raw IO with no address translation).

---

Changes from v2:

Finn Thain:
- add back stable Cc:

Changes from v1:

Damien Le Moal:
- change patch title
- drop stable backport tag

Changes from RFC v3:

- split off byte swap option into separate patch

Geert Uytterhoeven:
- review comments

Changes from RFC v2:
- add driver parameter 'data_swap' as bit mask for drives to swap

Changes from RFC v1:

Finn Thain:
- take care to supply IO address suitable for ioread8/iowrite8
- use MMIO address for data transfer

---
 drivers/ata/pata_falcon.c | 55 ++++++++++++++++++++++++---------------
 1 file changed, 34 insertions(+), 21 deletions(-)
drivers/ata/pata_falcon.c +34 -21
index 996516e64f13..346259e3bbc8 100644
@@ -123,8 +123,8 @@ static int __init pata_falcon_init_one(struct platform_device *pdev)
 	struct resource *base_res, *ctl_res, *irq_res;
 	struct ata_host *host;
 	struct ata_port *ap;
-	void __iomem *base;
-	int irq = 0;
+	void __iomem *base, *ctl_base;
+	int irq = 0, io_offset = 1, reg_scale = 4;
 
 	dev_info(&pdev->dev, "Atari Falcon and Q40/Q60 PATA controller\n");
 
@@ -165,26 +165,39 @@ static int __init pata_falcon_init_one(struct platform_device *pdev)
 	ap->pio_mask = ATA_PIO4;
 	ap->flags |= ATA_FLAG_SLAVE_POSS | ATA_FLAG_NO_IORDY;
 
-	base = (void __iomem *)base_mem_res->start;
 	/* N.B. this assumes data_addr will be used for word-sized I/O only */
-	ap->ioaddr.data_addr		= base + 0 + 0 * 4;
-	ap->ioaddr.error_addr		= base + 1 + 1 * 4;
-	ap->ioaddr.feature_addr		= base + 1 + 1 * 4;
-	ap->ioaddr.nsect_addr		= base + 1 + 2 * 4;
-	ap->ioaddr.lbal_addr		= base + 1 + 3 * 4;
-	ap->ioaddr.lbam_addr		= base + 1 + 4 * 4;
-	ap->ioaddr.lbah_addr		= base + 1 + 5 * 4;
-	ap->ioaddr.device_addr		= base + 1 + 6 * 4;
-	ap->ioaddr.status_addr		= base + 1 + 7 * 4;
-	ap->ioaddr.command_addr		= base + 1 + 7 * 4;
-
-	base = (void __iomem *)ctl_mem_res->start;
-	ap->ioaddr.altstatus_addr	= base + 1;
-	ap->ioaddr.ctl_addr		= base + 1;
-
-	ata_port_desc(ap, "cmd 0x%lx ctl 0x%lx",
-		      (unsigned long)base_mem_res->start,
-		      (unsigned long)ctl_mem_res->start);
+	ap->ioaddr.data_addr = (void __iomem *)base_mem_res->start;
+
+	if (base_res) {		/* only Q40 has IO resources */
+		io_offset = 0x10000;
+		reg_scale = 1;
+		base = (void __iomem *)base_res->start;
+		ctl_base = (void __iomem *)ctl_res->start;
+
+		ata_port_desc(ap, "cmd %pa ctl %pa",
+			      &base_res->start,
+			      &ctl_res->start);
+	} else {
+		base = (void __iomem *)base_mem_res->start;
+		ctl_base = (void __iomem *)ctl_mem_res->start;
+
+		ata_port_desc(ap, "cmd %pa ctl %pa",
+			      &base_mem_res->start,
+			      &ctl_mem_res->start);
+	}
+
+	ap->ioaddr.error_addr	= base + io_offset + 1 * reg_scale;
+	ap->ioaddr.feature_addr	= base + io_offset + 1 * reg_scale;
+	ap->ioaddr.nsect_addr	= base + io_offset + 2 * reg_scale;
+	ap->ioaddr.lbal_addr	= base + io_offset + 3 * reg_scale;
+	ap->ioaddr.lbam_addr	= base + io_offset + 4 * reg_scale;
+	ap->ioaddr.lbah_addr	= base + io_offset + 5 * reg_scale;
+	ap->ioaddr.device_addr	= base + io_offset + 6 * reg_scale;
+	ap->ioaddr.status_addr	= base + io_offset + 7 * reg_scale;
+	ap->ioaddr.command_addr	= base + io_offset + 7 * reg_scale;
+
+	ap->ioaddr.altstatus_addr	= ctl_base + io_offset;
+	ap->ioaddr.ctl_addr		= ctl_base + io_offset;
 
 	irq_res = platform_get_resource(pdev, IORESOURCE_IRQ, 0);
 	if (irq_res && irq_res->start > 0) {
Signature · 1 line
2.17.1
#2 lore raw

Re: [PATCH v3 1/2] ata: pata_falcon: fix IO base selection for Q40

Hi Michael,

Quoted 29 lines · Michael Schmitz

With commit 44b1fbc0f5f3 ("m68k/q40: Replace q40ide driver with pata_falcon and falconide"), the Q40 IDE driver was replaced by pata_falcon.c.

Both IO and memory resources were defined for the Q40 IDE platform device, but definition of the IDE register addresses was modeled after the Falcon case, both in use of the memory resources and in including register scale and byte vs. word offset in the address.

This was correct for the Falcon case, which does not apply any address translation to the register addresses. In the Q40 case, all of device base address, byte access offset and register scaling is included in the platform specific ISA access translation (in asm/mm_io.h).

As a consequence, such address translation gets applied twice, and register addresses are mangled.

Use the device base address from the platform IO resource, and use standard register offsets from that base in order to calculate register addresses (the IO address translation will then apply the correct ISA window base and scaling).

Encode PIO_OFFSET into IO port addresses for all registers except the data transfer register. Encode the MMIO offset there (pata_falcon_data_xfer() directly uses raw IO with no address translation).

Quoted 9 lines

Thanks for the update!

Quoted 27 lines
drivers/ata/pata_falcon.c +1 -19
@@ -165,26 +165,39 @@ static int __init pata_falcon_init_one(struct platform_device *pdev)
        ap->pio_mask = ATA_PIO4;
        ap->flags |= ATA_FLAG_SLAVE_POSS | ATA_FLAG_NO_IORDY;
 
-       base = (void __iomem *)base_mem_res->start;
        /* N.B. this assumes data_addr will be used for word-sized I/O only */
-       ap->ioaddr.data_addr            = base + 0 + 0 * 4;
-       ap->ioaddr.error_addr           = base + 1 + 1 * 4;
-       ap->ioaddr.feature_addr         = base + 1 + 1 * 4;
-       ap->ioaddr.nsect_addr           = base + 1 + 2 * 4;
-       ap->ioaddr.lbal_addr            = base + 1 + 3 * 4;
-       ap->ioaddr.lbam_addr            = base + 1 + 4 * 4;
-       ap->ioaddr.lbah_addr            = base + 1 + 5 * 4;
-       ap->ioaddr.device_addr          = base + 1 + 6 * 4;
-       ap->ioaddr.status_addr          = base + 1 + 7 * 4;
-       ap->ioaddr.command_addr         = base + 1 + 7 * 4;
-
-       base = (void __iomem *)ctl_mem_res->start;
-       ap->ioaddr.altstatus_addr       = base + 1;
-       ap->ioaddr.ctl_addr             = base + 1;
-
-       ata_port_desc(ap, "cmd 0x%lx ctl 0x%lx",
-                     (unsigned long)base_mem_res->start,
-                     (unsigned long)ctl_mem_res->start);
+       ap->ioaddr.data_addr = (void __iomem *)base_mem_res->start;

drivers/ata/pata_falcon.c static int __init pata_falcon_init_one(struct platform_device *pdev) · line 179

Quoted 10 lines
drivers/ata/pata_falcon.c +10 -0
+
+       if (base_res) {         /* only Q40 has IO resources */
+               io_offset = 0x10000;
+               reg_scale = 1;
+               base = (void __iomem *)base_res->start;
+               ctl_base = (void __iomem *)ctl_res->start;
+
+               ata_port_desc(ap, "cmd %pa ctl %pa",
+                             &base_res->start,
+                             &ctl_res->start);

This can be moved outside the else, using %px to format base and ctl_base.

drivers/ata/pata_falcon.c static int __init pata_falcon_init_one(struct platform_device *pdev) · line 187

Quoted 8 lines
drivers/ata/pata_falcon.c +8 -0
+       } else {
+               base = (void __iomem *)base_mem_res->start;
+               ctl_base = (void __iomem *)ctl_mem_res->start;
+
+               ata_port_desc(ap, "cmd %pa ctl %pa",
+                             &base_mem_res->start,
+                             &ctl_mem_res->start);
+       }

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
#3 lore raw

Re: [PATCH v3 1/2] ata: pata_falcon: fix IO base selection for Q40

Hi Geert,

Quoted 74 lines · Geert Uytterhoeven

Hi Michael,

Quoted 29 lines · Michael Schmitz

With commit 44b1fbc0f5f3 ("m68k/q40: Replace q40ide driver with pata_falcon and falconide"), the Q40 IDE driver was replaced by pata_falcon.c.

Both IO and memory resources were defined for the Q40 IDE platform device, but definition of the IDE register addresses was modeled after the Falcon case, both in use of the memory resources and in including register scale and byte vs. word offset in the address.

This was correct for the Falcon case, which does not apply any address translation to the register addresses. In the Q40 case, all of device base address, byte access offset and register scaling is included in the platform specific ISA access translation (in asm/mm_io.h).

As a consequence, such address translation gets applied twice, and register addresses are mangled.

Use the device base address from the platform IO resource, and use standard register offsets from that base in order to calculate register addresses (the IO address translation will then apply the correct ISA window base and scaling).

Encode PIO_OFFSET into IO port addresses for all registers except the data transfer register. Encode the MMIO offset there (pata_falcon_data_xfer() directly uses raw IO with no address translation).

Quoted 9 lines

Thanks for the update!

Quoted 31 lines
drivers/ata/pata_falcon.c +5 -19
@@ -165,26 +165,39 @@ static int __init pata_falcon_init_one(struct platform_device *pdev)
         ap->pio_mask = ATA_PIO4;
         ap->flags |= ATA_FLAG_SLAVE_POSS | ATA_FLAG_NO_IORDY;
 
-       base = (void __iomem *)base_mem_res->start;
         /* N.B. this assumes data_addr will be used for word-sized I/O only */
-       ap->ioaddr.data_addr            = base + 0 + 0 * 4;
-       ap->ioaddr.error_addr           = base + 1 + 1 * 4;
-       ap->ioaddr.feature_addr         = base + 1 + 1 * 4;
-       ap->ioaddr.nsect_addr           = base + 1 + 2 * 4;
-       ap->ioaddr.lbal_addr            = base + 1 + 3 * 4;
-       ap->ioaddr.lbam_addr            = base + 1 + 4 * 4;
-       ap->ioaddr.lbah_addr            = base + 1 + 5 * 4;
-       ap->ioaddr.device_addr          = base + 1 + 6 * 4;
-       ap->ioaddr.status_addr          = base + 1 + 7 * 4;
-       ap->ioaddr.command_addr         = base + 1 + 7 * 4;
-
-       base = (void __iomem *)ctl_mem_res->start;
-       ap->ioaddr.altstatus_addr       = base + 1;
-       ap->ioaddr.ctl_addr             = base + 1;
-
-       ata_port_desc(ap, "cmd 0x%lx ctl 0x%lx",
-                     (unsigned long)base_mem_res->start,
-                     (unsigned long)ctl_mem_res->start);
+       ap->ioaddr.data_addr = (void __iomem *)base_mem_res->start;
+
+       if (base_res) {         /* only Q40 has IO resources */
+               io_offset = 0x10000;
+               reg_scale = 1;
Quoted 8 lines

drivers/ata/pata_falcon.c static int __init pata_falcon_init_one(struct platform_device *pdev) · line 179

Quoted 6 lines
drivers/ata/pata_falcon.c +6 -0
+               base = (void __iomem *)base_res->start;
+               ctl_base = (void __iomem *)ctl_res->start;
+
+               ata_port_desc(ap, "cmd %pa ctl %pa",
+                             &base_res->start,
+                             &ctl_res->start);

This can be moved outside the else, using %px to format base and ctl_base.

Right - do we need some additional message spelling out what address Q40 uses for data transfers? (Redundant for Falcon, of course ...)

Though that could be handled outside the else, too:

ata_port_desc(ap, "cmd %px ctl %px data %pa",

              base, ctl_base, &ap->ioaddr.data_addr);

Cheers,

    Michael

Quoted 12 lines

drivers/ata/pata_falcon.c static int __init pata_falcon_init_one(struct platform_device *pdev) · line 187

Quoted 8 lines
drivers/ata/pata_falcon.c +8 -0
+       } else {
+               base = (void __iomem *)base_mem_res->start;
+               ctl_base = (void __iomem *)ctl_mem_res->start;
+
+               ata_port_desc(ap, "cmd %pa ctl %pa",
+                             &base_mem_res->start,
+                             &ctl_mem_res->start);
+       }

Gr{oetje,eeting}s,

Geert

#4 lore raw

Re: [PATCH v3 1/2] ata: pata_falcon: fix IO base selection for Q40

Hi Michael,

Quoted 82 lines · Michael Schmitz
Quoted 72 lines · Geert Uytterhoeven
Quoted 29 lines · Michael Schmitz

With commit 44b1fbc0f5f3 ("m68k/q40: Replace q40ide driver with pata_falcon and falconide"), the Q40 IDE driver was replaced by pata_falcon.c.

Both IO and memory resources were defined for the Q40 IDE platform device, but definition of the IDE register addresses was modeled after the Falcon case, both in use of the memory resources and in including register scale and byte vs. word offset in the address.

This was correct for the Falcon case, which does not apply any address translation to the register addresses. In the Q40 case, all of device base address, byte access offset and register scaling is included in the platform specific ISA access translation (in asm/mm_io.h).

As a consequence, such address translation gets applied twice, and register addresses are mangled.

Use the device base address from the platform IO resource, and use standard register offsets from that base in order to calculate register addresses (the IO address translation will then apply the correct ISA window base and scaling).

Encode PIO_OFFSET into IO port addresses for all registers except the data transfer register. Encode the MMIO offset there (pata_falcon_data_xfer() directly uses raw IO with no address translation).

Quoted 9 lines

Thanks for the update!

Quoted 31 lines
drivers/ata/pata_falcon.c +5 -19
@@ -165,26 +165,39 @@ static int __init pata_falcon_init_one(struct platform_device *pdev)
         ap->pio_mask = ATA_PIO4;
         ap->flags |= ATA_FLAG_SLAVE_POSS | ATA_FLAG_NO_IORDY;
 
-       base = (void __iomem *)base_mem_res->start;
         /* N.B. this assumes data_addr will be used for word-sized I/O only */
-       ap->ioaddr.data_addr            = base + 0 + 0 * 4;
-       ap->ioaddr.error_addr           = base + 1 + 1 * 4;
-       ap->ioaddr.feature_addr         = base + 1 + 1 * 4;
-       ap->ioaddr.nsect_addr           = base + 1 + 2 * 4;
-       ap->ioaddr.lbal_addr            = base + 1 + 3 * 4;
-       ap->ioaddr.lbam_addr            = base + 1 + 4 * 4;
-       ap->ioaddr.lbah_addr            = base + 1 + 5 * 4;
-       ap->ioaddr.device_addr          = base + 1 + 6 * 4;
-       ap->ioaddr.status_addr          = base + 1 + 7 * 4;
-       ap->ioaddr.command_addr         = base + 1 + 7 * 4;
-
-       base = (void __iomem *)ctl_mem_res->start;
-       ap->ioaddr.altstatus_addr       = base + 1;
-       ap->ioaddr.ctl_addr             = base + 1;
-
-       ata_port_desc(ap, "cmd 0x%lx ctl 0x%lx",
-                     (unsigned long)base_mem_res->start,
-                     (unsigned long)ctl_mem_res->start);
+       ap->ioaddr.data_addr = (void __iomem *)base_mem_res->start;
+
+       if (base_res) {         /* only Q40 has IO resources */
+               io_offset = 0x10000;
+               reg_scale = 1;
Quoted 8 lines

drivers/ata/pata_falcon.c static int __init pata_falcon_init_one(struct platform_device *pdev) · line 179

Quoted 6 lines
drivers/ata/pata_falcon.c +6 -0
+               base = (void __iomem *)base_res->start;
+               ctl_base = (void __iomem *)ctl_res->start;
+
+               ata_port_desc(ap, "cmd %pa ctl %pa",
+                             &base_res->start,
+                             &ctl_res->start);

This can be moved outside the else, using %px to format base and ctl_base.

Quoted 7 lines

Right - do we need some additional message spelling out what address Q40 uses for data transfers? (Redundant for Falcon, of course ...)

Though that could be handled outside the else, too:

ata_port_desc(ap, "cmd %px ctl %px data %pa",

               base, ctl_base, &ap->ioaddr.data_addr);

I guess that wouldn't hurt.

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
#5 lore raw

Re: [PATCH v3 1/2] ata: pata_falcon: fix IO base selection for Q40

Hi Geert,

Am 22.08.2023 um 19:05 schrieb Geert Uytterhoeven:

Quoted 15 lines
Quoted 15 lines
Quoted 11 lines

drivers/ata/pata_falcon.c static int __init pata_falcon_init_one(struct platform_device *pdev) · line 179

Quoted 9 lines
drivers/ata/pata_falcon.c +9 -0
+       if (base_res) {         /* only Q40 has IO resources */
+               io_offset = 0x10000;
+               reg_scale = 1;
+               base = (void __iomem *)base_res->start;
+               ctl_base = (void __iomem *)ctl_res->start;
+
+               ata_port_desc(ap, "cmd %pa ctl %pa",
+                             &base_res->start,
+                             &ctl_res->start);

This can be moved outside the else, using %px to format base and ctl_base.

Right - do we need some additional message spelling out what address Q40 uses for data transfers? (Redundant for Falcon, of course ...)

Quoted 6 lines
Quoted 4 lines

Though that could be handled outside the else, too:

ata_port_desc(ap, "cmd %px ctl %px data %pa",

               base, ctl_base, &ap->ioaddr.data_addr);

I guess that wouldn't hurt.

Done - I'll send out v4 tomorrow.

Cheers,

Michael

Quoted 5 lines

Gr{oetje,eeting}s,

Geert

#6 lore raw

Re: [PATCH v3 1/2] ata: pata_falcon: fix IO base selection for Q40

In reply to Geert Uytterhoeven

Hi Geert,

Am 21.08.2023 um 19:50 schrieb Geert Uytterhoeven:

Quoted 6 lines
Quoted 6 lines
drivers/ata/pata_falcon.c +6 -0
+       ap->ioaddr.data_addr = (void __iomem *)base_mem_res->start;
+
+       if (base_res) {         /* only Q40 has IO resources */
+               io_offset = 0x10000;
+               reg_scale = 1;
+               base = (void __iomem *)base_res->start;
Quoted 8 lines

drivers/ata/pata_falcon.c static int __init pata_falcon_init_one(struct platform_device *pdev) · line 179

Quoted 5 lines
drivers/ata/pata_falcon.c +5 -0
+               ctl_base = (void __iomem *)ctl_res->start;
+
+               ata_port_desc(ap, "cmd %pa ctl %pa",
+                             &base_res->start,
+                             &ctl_res->start);

This can be moved outside the else, using %px to format base and ctl_base.

I get a checkpatch warning for %px, but not for %pa (used for . &ap->ioaddr.data_addr). What gives?

WARNING: Using vsprintf specifier '%px' potentially exposes the kernel memory layout, if you don't really need the address please consider using '%p'. #148: FILE: drivers/ata/pata_falcon.c:194:
+ ata_port_desc(ap, "cmd %px ctl %px data %pa",
+ base, ctl_base, &ap->ioaddr.data_addr);

Using %pa and &base, &ctl_base just to shut that up seems a little silly ...

Cheers,

Michael

Quoted 14 lines

drivers/ata/pata_falcon.c static int __init pata_falcon_init_one(struct platform_device *pdev) · line 187

Quoted 9 lines
drivers/ata/pata_falcon.c +8 -0
+       } else {
+               base = (void __iomem *)base_mem_res->start;
+               ctl_base = (void __iomem *)ctl_mem_res->start;
+
+               ata_port_desc(ap, "cmd %pa ctl %pa",
+                             &base_mem_res->start,
+                             &ctl_mem_res->start);
+       }

Gr{oetje,eeting}s,

Geert

#7 lore raw

Re: [PATCH v3 1/2] ata: pata_falcon: fix IO base selection for Q40

Hi Michael,

Quoted 23 lines · Michael Schmitz

Am 21.08.2023 um 19:50 schrieb Geert Uytterhoeven:

Quoted 6 lines
Quoted 6 lines
drivers/ata/pata_falcon.c +6 -0
+       ap->ioaddr.data_addr = (void __iomem *)base_mem_res->start;
+
+       if (base_res) {         /* only Q40 has IO resources */
+               io_offset = 0x10000;
+               reg_scale = 1;
+               base = (void __iomem *)base_res->start;
Quoted 8 lines

drivers/ata/pata_falcon.c static int __init pata_falcon_init_one(struct platform_device *pdev) · line 179

Quoted 5 lines
drivers/ata/pata_falcon.c +5 -0
+               ctl_base = (void __iomem *)ctl_res->start;
+
+               ata_port_desc(ap, "cmd %pa ctl %pa",
+                             &base_res->start,
+                             &ctl_res->start);

This can be moved outside the else, using %px to format base and ctl_base.

I get a checkpatch warning for %px, but not for %pa (used for . &ap->ioaddr.data_addr). What gives?

WARNING: Using vsprintf specifier '%px' potentially exposes the kernel memory layout, if you don't really need the address please consider using '%p'.
#148: FILE: drivers/ata/pata_falcon.c:194:

Review

Quoted 2 lines
(quoted diff) +2 -0
+       ata_port_desc(ap, "cmd %px ctl %px data %pa",
+                     base, ctl_base, &ap->ioaddr.data_addr);

You can ignore this.
%p prints obfuscated pointer values, so they are useless here. %px prints an unobfuscated pointer value, which is fine here, as it's not a pointer to kernel memory that an attacker can use, but just the virtual address of an I/O device mapped one-to-one.

Gr{oetje,eeting}s,

Geert

Signature · 5 lines
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds
#8 lore raw

PATCH v3 2/2 ata: pata_falcon: add data_swab option to byte-swap disk data

In reply to Michael Schmitz

Some users of pata_falcon on Q40 have IDE disks in default IDE little endian byte order, whereas legacy disks use host-native big-endian byte order as on the Atari Falcon.

Add module parameter 'data_swab' to allow connecting drives with non-native data byte order. Drives selected by the data_swap bit mask will have their user data byte-swapped to host byte order, i.e. 'pata_falcon.data_swab=2' will byte-swap all user data on drive B, leaving data on drive A in native byte order. On Q40, drives on a second IDE interface may be added to the bit mask as bits 2 and 3.

Default setting is no byte swapping, i.e. compatibility with the native Falcon or Q40 operating system disk format.

Cc William R Sowerbutts <[email protected]>
Cc Finn Thain <[email protected]>
Cc Geert Uytterhoeven <[email protected]>

---

Changes since v2:

Geert Uytterhoeven:
- only shift swap bitmask if pdev->id > 0

Finn Thain:
- use pdev->devno directly for byte swap check

Changes since v1:

Damien Le Moal:
- change patch title
- drop swap_data flag

Finn Thain:
- drop allocation of ap->private struct, use field as bitmask

Changes since RFC v4:

Geert Uytterhoeven:
- don't shift static module parameter for drive 3/4 bitmask
- simplify bit mask calculation to always use pdev->id

Finn Thain:
- correct bit numbers for drive 3/4

Changes since RFC v3:

- split off this byte swap handling into separate patch

- add hint regarding third and fourth drive on Q40

Finn Thain:
- rename module parameter to 'data_swab' to better reflect its use

William Sowerbutts:
- correct IDE drive number used in data swap conditional

---
 drivers/ata/pata_falcon.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
drivers/ata/pata_falcon.c +9 -1
index 346259e3bbc8..e95658bc36e3 100644
@@ -33,6 +33,11 @@
 #define DRV_NAME "pata_falcon"
 #define DRV_VERSION "0.1.0"
 
+static int pata_falcon_swap_mask;
+
+module_param_named(data_swab, pata_falcon_swap_mask, int, 0444);
+MODULE_PARM_DESC(data_swab, "Data byte swap enable/disable bitmap (0x1==drive1, 0x2==drive2, 0x4==drive3, 0x8==drive4, default==0)");
+
 static const struct scsi_host_template pata_falcon_sht = {
 	ATA_PIO_SHT(DRV_NAME),
 };
@@ -50,7 +55,7 @@ static unsigned int pata_falcon_data_xfer(struct ata_queued_cmd *qc,
 
 	if (dev->class == ATA_DEV_ATA && cmd &&
 	    !blk_rq_is_passthrough(scsi_cmd_to_rq(cmd)))
-		swap = 0;
+		swap = (uintptr_t)ap->private_data & BIT(dev->devno);
 
 	/* Transfer multiple of 2 bytes */
 	if (rw == READ) {
@@ -199,6 +204,9 @@ static int __init pata_falcon_init_one(struct platform_device *pdev)
 	ap->ioaddr.altstatus_addr	= ctl_base + io_offset;
 	ap->ioaddr.ctl_addr		= ctl_base + io_offset;
 
+	ap->private_data = (void *)(uintptr_t)(pdev->id > 0 ?
+		pata_falcon_swap_mask >> 2 : pata_falcon_swap_mask);
+
 	irq_res = platform_get_resource(pdev, IORESOURCE_IRQ, 0);
 	if (irq_res && irq_res->start > 0) {
 		irq = irq_res->start;
Signature · 1 line
2.17.1