of: Fix comparison of "compatible" properties
PATCH 1/1 of: Fix comparison of "compatible" properties
Commit 7c7b60cb87547b1664a4385c187f029bf514a737
"of: put default string compare and #a/s-cell values into common header"
Breaks various things on powerpc due to using strncasecmp instead of strcasecmp for comparing against "compatible" strings.
This causes things like the 4xx PCI code to fail miserably due to the partial matches in code like this:
for_each_compatible_node(np, NULL, "ibm,plb-pcix")
ppc4xx_probe_pcix_bridge(np);
for_each_compatible_node(np, NULL, "ibm,plb-pci")
ppc4xx_probe_pci_bridge(np);
This reverts us to use strcasecmp. I do wonder why microblase and sparc want the partial matches though. For sparc it could be historical, but microblaze is probably doing the wrong thing so this patch also changes the microblaze behaviour to use not allow partial matches. Michal will surely beat me over the head with a cluestick if I'm wrong, in which case I'll fix this patch.
It's not quite right to do partial name match. Entries in a compatible list are meant to be matched whole. If a device is compatible with both "foo" and "foo1", then the device should have both strings in its "compatible" property.
(for patch description)
---
include/linux/of.h | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
include/linux/of.h +1 -1
@@ -127,7 +127,7 @@ static inline unsigned long of_read_ulong(const __be32 *cell, int size)/* Default string compare functions, Allow arch asm/prom.h to override */ #if !defined(of_compat_cmp) -#define of_compat_cmp (s1, s2, l) strncasecmp((s1), (s2), (l)) +#define of_compat_cmp (s1, s2, l) strcasecmp((s1), (s2))#define of_prop_cmp (s1, s2) strcmp((s1), (s2))#define of_node_cmp (s1, s2) strcasecmp((s1), (s2))#endif
From: Grant Likely <grant.likely-s3s/[email protected]>
Date: Wed, 17 Mar 2010 19:01:00 -0600
Quoted 21 lines
Commit
7c7b60cb87547b1664a4385c187f029bf514a737
"of: put default string compare and #a/s-cell values into common header"Breaks various things on powerpc due to using strncasecmp instead of strcasecmp for comparing against "compatible" strings.
This causes things like the 4xx PCI code to fail miserably due to the partial matches in code like this:
for_each_compatible_node(np, NULL, "ibm,plb-pcix")
ppc4xx_probe_pcix_bridge(np);
for_each_compatible_node(np, NULL, "ibm,plb-pci")
ppc4xx_probe_pci_bridge(np);This reverts us to use strcasecmp. I do wonder why microblase and sparc want the partial matches though. For sparc it could be historical, but microblaze is probably doing the wrong thing so this patch also changes the microblaze behaviour to use not allow partial matches. Michal will surely beat me over the head with a cluestick if I'm wrong, in which case I'll fix this patch.
Quoted 8 lines
It's not quite right to do partial name match. Entries in a compatible list are meant to be matched whole. If a device is compatible with both "foo" and "foo1", then the device should have both strings in its "compatible" property.
Signed-off-by Benjamin Herrenschmidt <[email protected]>(for patch description)
Signed-off-by Grant Likely <grant.likely-s3s/[email protected]>
A message that is not in the archive (sent off-list or not yet received): 20100317.231847.112586488.davem-fT/PcQaiUtIeIZ0/[email protected]
Quoted 4 lines · David Miller
Quoted 1 line
It's not quite right to do partial name match. Entries in a
compatible
Quoted 1 line
list are meant to be matched whole. If a device is compatible with
both
Quoted 8 lines
Quoted 6 lines
"foo" and "foo1", then the device should have both strings in its "compatible" property.
Signed-off-by Benjamin Herrenschmidt <[email protected]>(for patch description)
Signed-off-by Grant Likely <grant.likely-s3s/[email protected]>
BTW. While at it... sparc still has it's override here to do partial match (strncmp iirc). Do you rely on it ? Or do you do that only to be case insensitive and the "n" is a result of history ?
Cheers,
Ben
Quoted 14 lines · Grant Likely
Commit
7c7b60cb87547b1664a4385c187f029bf514a737
"of: put default string compare and #a/s-cell values into common header"Breaks various things on powerpc due to using strncasecmp instead of strcasecmp for comparing against "compatible" strings.
This causes things like the 4xx PCI code to fail miserably due to the partial matches in code like this:
for_each_compatible_node(np, NULL, "ibm,plb-pcix")
ppc4xx_probe_pcix_bridge(np);
for_each_compatible_node(np, NULL, "ibm,plb-pci")
ppc4xx_probe_pci_bridge(np);
Quoted 6 lines
This reverts us to use strcasecmp. I do wonder why microblase and sparc want the partial matches though. For sparc it could be historical, but microblaze is probably doing the wrong thing so this patch also changes the microblaze behaviour to use not allow partial matches. Michal will surely beat me over the head with a cluestick if I'm wrong, in which case I'll fix this patch.
No, you are correct. I don't want partial matches.
Quoted 5 lines
It's not quite right to do partial name match. Entries in a compatible list are meant to be matched whole. If a device is compatible with both "foo" and "foo1", then the device should have both strings in its "compatible" property.
Nice example. I have better one. Look at of_serial.c. ns16550a must be for partial matches before ns16550. :-)
{ .type = "serial", .compatible = "ns16550a", .data = (void
*)PORT_16550A, },
{ .type = "serial", .compatible = "ns16550", .data = (void *)PORT_16550, },
Quoted 26 lines
Signed-off-by Benjamin Herrenschmidt <[email protected]>(for patch description)
Signed-off-by Grant Likely <grant.likely-s3s/[email protected]>CC David Miller <davem-fT/PcQaiUtIeIZ0/[email protected]>CC devicetree-discuss-uLR06cmDAlY/[email protected]CC Michal Simek <[email protected]>---
include/linux/of.h | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)include/linux/of.h +1
-1@@ -127,7 +127,7 @@ static inline unsigned long of_read_ulong(const __be32 *cell, int size)/* Default string compare functions, Allow arch asm/prom.h to override */ #if !defined(of_compat_cmp) -#define of_compat_cmp (s1, s2, l) strncasecmp((s1), (s2), (l)) +#define of_compat_cmp (s1, s2, l) strcasecmp((s1), (s2))#define of_prop_cmp (s1, s2) strcmp((s1), (s2))#define of_node_cmp (s1, s2) strcasecmp((s1), (s2))#endif
Signature · 4 lines
Michal Simek, Ing. (M.Eng) w: www.monstr.eu p: +42-0-721842854 Maintainer of Linux kernel 2.6 Microblaze Linux - http://www.monstr.eu/fdt/ Microblaze U-BOOT custodian