Skip to contentSkip to conversation

pcmcia IDE broken in 2.6.18-rc1

DISCUSSION Pavel Machek 4 participants 9 messages Jul 8, 2006 – Jul 13, 2006

lore
Share & tools
lore
#0 lore raw

Hi!

When I insert the card, I get

pccard: PCMCIA card inserted into slot 0
cs: memory probe 0xe8000000-0xefffffff: excluding 0xe8000000-0xefffffff
cs: memory probe 0xc0200000-0xcfffffff: excluding 0xc0200000-0xc11fffff 0xc1a00000-0xc61fffff 0xc6a00000-0xc71fffff 0xc7a00000-0xc81fffff 0xc8a00000-0xc91fffff 0xc9a00000-0xca1fffff 0xcaa00000-0xcb1fffff 0xcba00000-0xcc1fffff 0xcca00000-0xcd1fffff 0xcda00000-0xce1fffff 0xcea00000-0xcf1fffff 0xcfa00000-0xd01fffff
pcmcia: registering new device pcmcia0.0
PM: Adding info for pcmcia:0.0
ide2: I/O resource 0xF887E00E-0xF887E00E not free.
ide2: ports already in use, skipping probe
ide2: I/O resource 0xF887E01E-0xF887E01E not free.
ide2: ports already in use, skipping probe
...

it ends with

ide-cs: ide_register() at 0xf999c000 & 0xf999c00e, irq 7 failed

:-(. Back to 2.6.17 once again, I'm afraid...
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#1 lore raw
Quoted 20 lines · On Sat, 8 Jul 2006 16:55:43 +0200 Pavel Machek <[email protected]> wrote:

Hi!

When I insert the card, I get

pccard: PCMCIA card inserted into slot 0
cs: memory probe 0xe8000000-0xefffffff: excluding 0xe8000000-0xefffffff
cs: memory probe 0xc0200000-0xcfffffff: excluding 0xc0200000-0xc11fffff 0xc1a00000-0xc61fffff 0xc6a00000-0xc71fffff 0xc7a00000-0xc81fffff 0xc8a00000-0xc91fffff 0xc9a00000-0xca1fffff 0xcaa00000-0xcb1fffff 0xcba00000-0xcc1fffff 0xcca00000-0xcd1fffff 0xcda00000-0xce1fffff 0xcea00000-0xcf1fffff 0xcfa00000-0xd01fffff
pcmcia: registering new device pcmcia0.0
PM: Adding info for pcmcia:0.0
ide2: I/O resource 0xF887E00E-0xF887E00E not free.
ide2: ports already in use, skipping probe
ide2: I/O resource 0xF887E01E-0xF887E01E not free.
ide2: ports already in use, skipping probe
...

Quoted 5 lines

it ends with

ide-cs: ide_register() at 0xf999c000 & 0xf999c00e, irq 7 failed

:-(. Back to 2.6.17 once again, I'm afraid...

Appears to be the same bug as http://lkml.org/lkml/2006/6/15/155

That debugging effort dried up at "Can you do some more tracing on hwif."

You're our only hope. Can you debug it a bit please?

#2 lore raw
Quoted 18 lines
Quoted 18 lines

When I insert the card, I get

pccard: PCMCIA card inserted into slot 0
cs: memory probe 0xe8000000-0xefffffff: excluding 0xe8000000-0xefffffff
cs: memory probe 0xc0200000-0xcfffffff: excluding 0xc0200000-0xc11fffff 0xc1a00000-0xc61fffff 0xc6a00000-0xc71fffff 0xc7a00000-0xc81fffff 0xc8a00000-0xc91fffff 0xc9a00000-0xca1fffff 0xcaa00000-0xcb1fffff 0xcba00000-0xcc1fffff 0xcca00000-0xcd1fffff 0xcda00000-0xce1fffff 0xcea00000-0xcf1fffff 0xcfa00000-0xd01fffff
pcmcia: registering new device pcmcia0.0
PM: Adding info for pcmcia:0.0
ide2: I/O resource 0xF887E00E-0xF887E00E not free.
ide2: ports already in use, skipping probe
ide2: I/O resource 0xF887E01E-0xF887E01E not free.
ide2: ports already in use, skipping probe
...

Quoted 7 lines
Quoted 5 lines

it ends with

ide-cs: ide_register() at 0xf999c000 & 0xf999c00e, irq 7 failed

:-(. Back to 2.6.17 once again, I'm afraid...

Appears to be the same bug as http://lkml.org/lkml/2006/6/15/155

As I am one of the probably few people who have some PCMCIA-IDE-drive (Sony PCGA-CD51/A), running 2.6.18-rc1 gives me no problems.

pccard: PCMCIA card inserted into slot 0
kernel: cs: memory probe 0xa0000000-0xa0ffffff: clean.
pcmcia: registering new device pcmcia0.0
Probind IDE interface ide2...
hde: TOSHIBA CD-ROM XM-7002Bc, ATAPI CD/DVD-ROM drive ide2 at 0x180-0x198,0x386 on irq 3
ide-cs: hde: Vpp = 0.0
pcmcia: Detected deprecated PCMCIA ioctl usage from process: hald

Seems fine. I can mount and read it w/o problems.

$ rpm -q pcmciautils
pcmciautils-012-11 (SUSE Linux 10.1)

CONFIG_ISA=y
# CONFIG_RESOURCES_64BIT is not set

Just 2¢, ask for more if needed :)

Jan Engelhardt

#3 lore raw
In reply to Pavel Machek

Ar Sad, 2006-07-08 am 16:55 +0200, ysgrifennodd Pavel Machek:

Quoted 4 lines

ide2: I/O resource 0xF887E00E-0xF887E00E not free.
ide2: ports already in use, skipping probe
ide2: I/O resource 0xF887E01E-0xF887E01E not free.
ide2: ports already in use, skipping probe

Looks like ioremap values not I/O ports. Probably the various IDE layer changes from 2.6.17-mm.

My first guess would be the PCMCIA layer changes to use mmio ports are not setting hwif->mmio (I think its ->mmio) to 2 and doing their own resource management.

Alan

#4 lore raw
Quoted 7 lines · On Sat, 08 Jul 2006 18:36:39 +0100 Alan Cox <[email protected]> wrote:

Ar Sad, 2006-07-08 am 16:55 +0200, ysgrifennodd Pavel Machek:

Quoted 4 lines

ide2: I/O resource 0xF887E00E-0xF887E00E not free.
ide2: ports already in use, skipping probe
ide2: I/O resource 0xF887E01E-0xF887E01E not free.
ide2: ports already in use, skipping probe

Quoted 7 lines

Looks like ioremap values not I/O ports. Probably the various IDE layer changes from 2.6.17-mm.

My first guess would be the PCMCIA layer changes to use mmio ports are not setting hwif->mmio (I think its ->mmio) to 2 and doing their own resource management.

5040cb8b7e61b7a03e8837920b9eb2c839bb1947 looks like a good one to try reverting.

commit 5040cb8b7e61b7a03e8837920b9eb2c839bb1947

Author: Thomas Kleffel <[email protected]>
Date:   Sun May 14 15:16:30 2006 +0200

[PATCH] pcmcia: Make ide_cs work with the memory space of CF-Cards if IO space is not available

This patch enables ide_cs to access CF-cards via their common memory rather than via their IO space.

drivers/ide/legacy/ide-cs.c +71 -10
index 602797a..b7e459e 100644
@@ -146,7 +146,16 @@ static void ide_detach(struct pcmcia_dev
     kfree(link->priv);
 } /* ide_detach */
 
-static int idecs_register(unsigned long io, unsigned long ctl, unsigned long irq, struct pcmcia_device *handle)
+static void idecs_mmio_fixup(ide_hwif_t *hwif)
+{
+	default_hwif_mmiops(hwif);
+	hwif->mmio = 2;
+
+	ide_undecoded_slave(hwif);
+}
+
+static int idecs_register(unsigned long io, unsigned long ctl,
+	unsigned long irq, struct pcmcia_device *handle, int is_mmio)
 {
     hw_regs_t hw;
     memset(&hw, 0, sizeof(hw));
@@ -154,7 +163,19 @@ static int idecs_register(unsigned long 
     hw.irq = irq;
     hw.chipset = ide_pci;
     hw.dev = &handle->dev;
-    return ide_register_hw_with_fixup(&hw, NULL, ide_undecoded_slave);
+
+    if(is_mmio)
+    	return ide_register_hw_with_fixup(&hw, NULL, idecs_mmio_fixup);
+    else
+        return ide_register_hw_with_fixup(&hw, NULL, ide_undecoded_slave);
+}
+
+void outb_io(unsigned char value, unsigned long port) {
+	outb(value, port);
+}
+
+void outb_mem(unsigned char value, unsigned long port) {
+	writeb(value, (void __iomem *) port);
 }
 
 /*======================================================================
@@ -180,7 +201,8 @@ static int ide_config(struct pcmcia_devi
     } *stk = NULL;
     cistpl_cftable_entry_t *cfg;
     int i, pass, last_ret = 0, last_fn = 0, hd, is_kme = 0;
-    unsigned long io_base, ctl_base;
+    unsigned long io_base, ctl_base, is_mmio, try_slave;
+    void (*my_outb)(unsigned char, unsigned long);
 
     DEBUG(0, "ide_config(0x%p)\n", link);
 
@@ -210,7 +232,7 @@ static int ide_config(struct pcmcia_devi
     /* Not sure if this is right... look up the current Vcc */
     CS_CHECK(GetConfigurationInfo, pcmcia_get_configuration_info(link, &stk->conf));
 
-    pass = io_base = ctl_base = 0;
+    pass = io_base = ctl_base = is_mmio = try_slave = 0;
     tuple.DesiredTuple = CISTPL_CFTABLE_ENTRY;
     tuple.Attributes = 0;
     CS_CHECK(GetFirstTuple, pcmcia_get_first_tuple(link, &tuple));
@@ -258,11 +280,45 @@ static int ide_config(struct pcmcia_devi
 			goto next_entry;
 		io_base = link->io.BasePort1;
 		ctl_base = link->io.BasePort1 + 0x0e;
+
+		if (io->win[0].len >= 0x20)
+			try_slave = 1;
+
 	    } else goto next_entry;
 	    /* If we've got this far, we're done */
 	    break;
 	}
 
+	if ((cfg->mem.nwin > 0) || (stk->dflt.mem.nwin > 0)) {
+	    win_req_t req;
+	    memreq_t map;
+	    cistpl_mem_t *mem = (cfg->mem.nwin) ? &cfg->mem : &stk->dflt.mem;
+
+	    if (mem->win[0].len < 16)
+	    	goto next_entry;
+
+	    req.Attributes = WIN_DATA_WIDTH_16|WIN_MEMORY_TYPE_CM;
+	    req.Attributes |= WIN_ENABLE;
+	    req.Base = mem->win[0].host_addr;
+	    req.Size = 0;
+
+	    req.AccessSpeed = 0;
+	    if (pcmcia_request_window(&link, &req, &link->win) != 0)
+		goto next_entry;
+	    map.Page = 0; map.CardOffset = mem->win[0].card_addr;
+	    if (pcmcia_map_mem_page(link->win, &map) != 0)
+		goto next_entry;
+
+      	    io_base = (unsigned long) ioremap(req.Base, req.Size);
+    	    ctl_base = io_base + 0x0e;
+    	    is_mmio = 1;
+
+    	    if (mem->win[0].len >= 0x20)
+    	    	try_slave = 1;
+
+	    break;
+	}
+
     next_entry:
 	if (cfg->flags & CISTPL_CFTABLE_DEFAULT)
 	    memcpy(&stk->dflt, cfg, sizeof(stk->dflt));
@@ -278,21 +334,26 @@ static int ide_config(struct pcmcia_devi
     CS_CHECK(RequestIRQ, pcmcia_request_irq(link, &link->irq));
     CS_CHECK(RequestConfiguration, pcmcia_request_configuration(link, &link->conf));
 
+    if(is_mmio)
+    	my_outb = outb_mem;
+    else
+    	my_outb = outb_io;
+
     /* disable drive interrupts during IDE probe */
-    outb(0x02, ctl_base);
+    my_outb(0x02, ctl_base);
 
     /* special setup for KXLC005 card */
     if (is_kme)
-	outb(0x81, ctl_base+1);
+	my_outb(0x81, ctl_base+1);
 
     /* retry registration in case device is still spinning up */
     for (hd = -1, i = 0; i < 10; i++) {
-	hd = idecs_register(io_base, ctl_base, link->irq.AssignedIRQ, link);
+	hd = idecs_register(io_base, ctl_base, link->irq.AssignedIRQ, link, is_mmio);
 	if (hd >= 0) break;
-	if (link->io.NumPorts1 == 0x20) {
-	    outb(0x02, ctl_base + 0x10);
+	if (try_slave) {
+	    my_outb(0x02, ctl_base + 0x10);
 	    hd = idecs_register(io_base + 0x10, ctl_base + 0x10,
-				link->irq.AssignedIRQ, link);
+				link->irq.AssignedIRQ, link, is_mmio);
 	    if (hd >= 0) {
 		io_base += 0x10;
 		ctl_base += 0x10;
#5 lore raw

Ar Sad, 2006-07-08 am 10:41 -0700, ysgrifennodd Andrew Morton:

drivers/ide/legacy/ide-cs.c static int ide_config(struct pcmcia_devi · line 314

Quoted 3 lines
drivers/ide/legacy/ide-cs.c +3 -0
+      	    io_base = (unsigned long) ioremap(req.Base, req.Size);
+    	    ctl_base = io_base + 0x0e;
+    	    is_mmio = 1;

Where does this get unmapped ?

#6 lore raw
Quoted 6 lines · On Sat, 08 Jul 2006 22:35:47 +0100 Alan Cox <[email protected]> wrote:

Ar Sad, 2006-07-08 am 10:41 -0700, ysgrifennodd Andrew Morton:

drivers/ide/legacy/ide-cs.c static int ide_config(struct pcmcia_devi · line 314

Quoted 3 lines
drivers/ide/legacy/ide-cs.c +3 -0
+      	    io_base = (unsigned long) ioremap(req.Base, req.Size);
+    	    ctl_base = io_base + 0x0e;
+    	    is_mmio = 1;

Where does this get unmapped ?

I don't think it does. (drivers/ide/legacy/ide-cs.c:312)

#7 lore raw
In reply to Andrew Morton

Hi!

Quoted 9 lines
Quoted 9 lines
Quoted 4 lines

ide2: I/O resource 0xF887E00E-0xF887E00E not free.
ide2: ports already in use, skipping probe
ide2: I/O resource 0xF887E01E-0xF887E01E not free.
ide2: ports already in use, skipping probe

Looks like ioremap values not I/O ports. Probably the various IDE layer changes from 2.6.17-mm.

Quoted 7 lines
Quoted 4 lines

My first guess would be the PCMCIA layer changes to use mmio ports are not setting hwif->mmio (I think its ->mmio) to 2 and doing their own resource management.

5040cb8b7e61b7a03e8837920b9eb2c839bb1947 looks like a good one to try reverting.

2.6.18-rc1-mm1 works okay. Is that enough, or do you want me to try reverting just this patch?
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#8 lore raw
Quoted 16 lines · On Mon, 10 Jul 2006 00:47:08 +0200 Pavel Machek <[email protected]> wrote:

Hi!

Quoted 14 lines
Quoted 13 lines
Quoted 4 lines

ide2: I/O resource 0xF887E00E-0xF887E00E not free.
ide2: ports already in use, skipping probe
ide2: I/O resource 0xF887E01E-0xF887E01E not free.
ide2: ports already in use, skipping probe

Looks like ioremap values not I/O ports. Probably the various IDE layer changes from 2.6.17-mm.

My first guess would be the PCMCIA layer changes to use mmio ports are not setting hwif->mmio (I think its ->mmio) to 2 and doing their own resource management.

Quoted 5 lines
Quoted 2 lines

5040cb8b7e61b7a03e8837920b9eb2c839bb1947 looks like a good one to try reverting.

2.6.18-rc1-mm1 works okay. Is that enough, or do you want me to try reverting just this patch?

Nope, that's fine, thanks. I think we can say that 5040cb8b7e61b7a03e8837920b9eb2c839bb1947 is busted.

Let's give Thomas and Dominik a few days to think about it before we do the deed..