Compat
Subsystems (3)
cover Compat
Hi Luis,
You will find in answer of this mail the first part of the compat upgrade I am using to support a network-oriented module on kernel from 2.6.15 to 2.6.36.
All this code is available on my repository on github:
https://github.com/lacombar/linux-compat.git local
This repository also contains the second part I will post later concerning a tentative to reduce namespace pollution when using the compat module. As it is mostly mechanical, I will start another thread with a single proof of concept patch and gather feedbacks :)
Comments welcome !
Thanks,
- Arnaud
Arnaud Lacombe (44):
compat/2.6.17: add header's skeleton
compat/2.6.17: add compat macro for NET_SKB_PAD' compat/2.6.18: add compat skeleton for the hwrng' API compat/2.6.18: add compat inline for netdev_alloc_skb() compat/2.6.19: fix comment
compat/2.6.19: add typedef for bool' and true'/`false' compat/2.6.19: rework compat macro to kmem_cache_destroy() compat/2.6.20: add header's skeleton
compat/2.6.20: add compat macro for `struct delayed_work' compat/2.6.23: add compat inline for notifier_{from,to}_errno() compat/2.6.24: add compat macro for sk_alloc() compat/2.6.25: don't include <linux/hw_random.h> prior to 2.6.18 compat/2.6.25: don't include <linux/{io,leds}.h> prior to 2.6.17 compat/2.6.25: conditionally defines `__WARN' compat/2.6.25: add compat macro for ip_route_output_key() compat/2.6.29: don't include <linux/if_link.h> prior to 2.6.19 compat/2.6.29: const'ify compat inline to dev_get_stats() compat/2.6.31: gather all the threaded IRQ backport together compat/2.6.31: fix threaded irq backport wrt. INIT_WORK() macro compat/2.6.32: fix compat inline to flush_delayed_work() compat/2.6.32: add compat macros for sysctl helpers compat/2.6.33: add compat macro to BUILD_BUG_ON_NOT_POWER_OF_2() compat/2.6.33: fix pcmcia include
compat/2.6.33: don't export prototype for pccard_loop_tuple() compat/2.6.36: `va_list' needs <stdarg.h>
compat: remove a couple of macro definition protection compat/2.6.33: remove macro not in mainline
compat: avoid using #include_next' directive in compat headers compat/2.6.22: rework uninitialized_var' and `__maybe_unused' macros compat/2.6.22: nuke pci_{clear,set}_mwi()
compat/2.6.22: add compat for couple of TCP accessors compat/2.6.22: add compat inline for ip_hdrlen() compat/2.6.22: constify skb_transport_header()'s argument compat/2.6.24: move 'struct net' to its own compat header compat/2.6.26: {U,}SHORT_MAX/SHORT_MIN have been deprecated compat/2.6.26: move unaligned backport to its own header compat/2.6.2{7,8}: fix definition for IEEE80211_HT_CAP_DSSSCCK40 compat/2.6.2{8,9}: fix tracepoint macros
compat/2.6.29: protect headers inclusion
compat/2.6.29: don't waste space around DIV_ROUND_CLOSEST() compat/2.6.29: fix typos
compat/2.6.{21,30}: make dev_set_uevent_suppress() a no-op for
<2.6.21
compat/2.6.{19,31}: fix threaded irq backport for <2.6.19 compat/2.6.35: 2.6.35 does not define `SHRT_MIN'
include/linux/asm/unaligned.h | 194 ++++++++++++++++++++++++++++++++++++ include/linux/compat-2.6.17.h | 13 +++ include/linux/compat-2.6.18.h | 45 +++++++++ include/linux/compat-2.6.19.h | 23 ++--- include/linux/compat-2.6.20.h | 14 +++ include/linux/compat-2.6.22.h | 41 ++++++-- include/linux/compat-2.6.23.h | 18 +++- include/linux/compat-2.6.24.h | 34 +------ include/linux/compat-2.6.25.h | 10 ++- include/linux/compat-2.6.26.h | 219 +---------------------------------------- include/linux/compat-2.6.27.h | 3 - include/linux/compat-2.6.28.h | 27 +----- include/linux/compat-2.6.29.h | 40 ++++++-- include/linux/compat-2.6.30.h | 2 + include/linux/compat-2.6.31.h | 54 ++++++++--- include/linux/compat-2.6.32.h | 31 ++++++- include/linux/compat-2.6.33.h | 12 +-- include/linux/compat-2.6.35.h | 1 - include/linux/compat-2.6.36.h | 1 + include/linux/pm_qos_params.h | 7 +- include/linux/tracepoint.h | 14 ++-- include/net/net_namespace.h | 50 ++++++++-- include/trace/define_trace.h | 5 - 23 files changed, 503 insertions(+), 355 deletions(-) create mode 100644 include/linux/asm/unaligned.h create mode 100644 include/linux/compat-2.6.17.h create mode 100644 include/linux/compat-2.6.20.h
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 1/44 compat/2.6.17: add header's skeleton
--- include/linux/compat-2.6.17.h | 11 +++++++++++ 1 files changed, 11 insertions(+), 0 deletions(-) create mode 100644 include/linux/compat-2.6.17.h
include/linux/compat-2.6.17.h new file +11 -0
@@ -0,0 +1,11 @@ +#ifndef LINUX_26_17_COMPAT_H +#define LINUX_26_17_COMPAT_H + +#include <linux/version.h> + +/* Compat work for 2.6.17 */ +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,17 )) + +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,17)) */ + +#endif /* LINUX_26_17_COMPAT_H */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 2/44 compat/2.6.17: add compat macro for `NET_SKB_PAD'
In reply to Arnaud Lacombe--- include/linux/compat-2.6.17.h | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/compat-2.6.17.h +2 -0
@@ -6,6 +6,8 @@/* Compat work for 2.6.17 */ #if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,17 )) +#define NET_SKB_PAD 16 +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,17)) */ #endif /* LINUX_26_17_COMPAT_H */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 3/44 compat/2.6.18: add compat skeleton for the `hwrng' API
In reply to Arnaud Lacombe--- include/linux/compat-2.6.18.h | 24 ++++++++++++++++++++++++ 1 files changed, 24 insertions(+), 0 deletions(-)
include/linux/compat-2.6.18.h +24 -0
@@ -8,6 +8,30 @@#define roundup (x, y) ((((x) + ((y) - 1)) / (y)) * (y)) +struct hwrng +{ +const char *name; +int (*init)(struct hwrng *rng); +void (*cleanup)(struct hwrng *rng); +int (*data_present)(struct hwrng *rng); +int (*data_read)(struct hwrng *rng,u32 *data); +unsigned long priv; + +/* internal. */ +struct list_head list; +}; + +static inline int +hwrng_register (struct hwrng *rng) +{ +return -ENOSYS ; +} + +static void +hwrng_unregister (struct hwrng *rng) +{ +} +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18)) */ #endif /* LINUX_26_18_COMPAT_H */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 4/44 compat/2.6.18: add compat inline for netdev_alloc_skb()
In reply to Arnaud Lacombe--- include/linux/compat-2.6.18.h | 21 +++++++++++++++++++++ 1 files changed, 21 insertions(+), 0 deletions(-)
include/linux/compat-2.6.18.h +21 -0
@@ -6,8 +6,29 @@/* Compat work for 2.6.18 */ #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18)) +#include <linux/gfp.h> +#include <linux/netdevice.h> +#include <linux/skbuff.h> + #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y)) +#ifndef NET_SKB_PAD +#define NET_SKB_PAD 16 +#endif + +static inline struct sk_buff * +netdev_alloc_skb(struct net_device *dev, unsigned int length) +{ + struct sk_buff *skb; + + skb = alloc_skb(length + NET_SKB_PAD, GFP_ATOMIC); + if (likely(skb)) { + skb_reserve(skb, NET_SKB_PAD); + skb->dev = dev; + } + return skb; +} + struct hwrng { const char *name;
Signature · 1 line
1.7.2.30.gc37d7.dirty
Re: [PATCH 04/44] compat/2.6.18: add compat inline for netdev_alloc_skb()
Quoted 11 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>--- include/linux/compat-2.6.18.h | 21 +++++++++++++++++++++ 1 files changed, 21 insertions(+), 0 deletions(-)include/linux/compat-2.6.18.h +0
-0@@ -6,8 +6,29 @@/* Compat work for 2.6.18 */
include/linux/compat-2.6.18.h line 16
Quoted 10 lines
include/linux/compat-2.6.18.h +6
-0#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,18 )) +#include <linux/gfp.h> +#include <linux/netdevice.h> +#include <linux/skbuff.h> +#define roundup (x, y) ((((x) + ((y) - 1)) / (y)) * (y)) +#ifndef NET_SKB_PAD +#define NET_SKB_PAD 16
Why are you adding the ifndef for NET_SKB_PAD if you know you already added it for 2.6.17 on your patch 2/44 ?
Luis
Re: [PATCH 04/44] compat/2.6.18: add compat inline for netdev_alloc_skb()
Hi,
Quoted 20 lines · Luis R. Rodriguez
Quoted 19 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>---
include/linux/compat-2.6.18.h | 21 +++++++++++++++++++++1 files changed, 21 insertions(+), 0 deletions(-)include/linux/compat-2.6.18.h +4
-0@@ -6,8 +6,29 @@ /* Compat work for 2.6.18 */ #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18)) +#include <linux/gfp.h> +#include <linux/netdevice.h> +#include <linux/skbuff.h> + #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
Quoted 6 lines
include/linux/compat-2.6.18.h line 16
Quoted 2 lines
include/linux/compat-2.6.18.h +2
-0+#ifndef NET_SKB_PAD +#define NET_SKB_PAD 16Why are you adding the ifndef for
NET_SKB_PADif you know you already added it for 2.6.17 on your patch 2/44 ?
I should have lost a track when I made that change, it correctly appeared in 2.6.17. My mistake :)
- Arnaud
PATCH 5/44 compat/2.6.19: fix comment
In reply to Arnaud Lacombe--- include/linux/compat-2.6.19.h | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
include/linux/compat-2.6.19.h +1 -1
@@ -19,6 +19,6 @@ compat_kmem_cache_destroy(struct kmem_cache *cachep)#endif } -#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,19)) */ +#endif /* (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,19)) */ #endif /* LINUX_26_19_COMPAT_H */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 6/44 compat/2.6.19: add typedef for `bool' and `true'/`false'
In reply to Arnaud Lacombe--- include/linux/compat-2.6.19.h | 8 ++++++++ 1 files changed, 8 insertions(+), 0 deletions(-)
include/linux/compat-2.6.19.h +8 -0
@@ -19,6 +19,14 @@ compat_kmem_cache_destroy(struct kmem_cache *cachep)#endif } +#else /* (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,19)) */ + +typedef _Bool bool ; +enum { + false =0 , + true =1 +}; +#endif /* (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,19)) */ #endif /* LINUX_26_19_COMPAT_H */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 7/44 compat/2.6.19: rework compat macro to kmem_cache_destroy()
In reply to Arnaud Lacombe--- include/linux/compat-2.6.19.h | 11 +---------- 1 files changed, 1 insertions(+), 10 deletions(-)
include/linux/compat-2.6.19.h +1 -10
@@ -8,16 +8,7 @@#include <linux/slab.h> -static inline int -compat_kmem_cache_destroy (struct kmem_cache *cachep) -{ -#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,19 )) -return kmem_cache_destroy (cachep); -#else -kmem_cache_destroy (cachep); -return 0 ; -#endif -} +#define kmem_cache_destroy (cachep) ({ kmem_cache_destroy(cachep); 0; })#else /* (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,19)) */
Signature · 1 line
1.7.2.30.gc37d7.dirty
Re: [PATCH 07/44] compat/2.6.19: rework compat macro to kmem_cache_destroy()
Quoted 14 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>--- include/linux/compat-2.6.19.h | 11 +---------- 1 files changed, 1 insertions(+), 10 deletions(-)include/linux/compat-2.6.19.h +0
-1@@ -8,16 +8,7 @@#include <linux/slab.h> -static inline int
include/linux/compat-2.6.19.h line 11
Quoted 10 lines
include/linux/compat-2.6.19.h +1
-9-compat_kmem_cache_destroy(struct kmem_cache *cachep) -{ -#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,19 )) -return kmem_cache_destroy (cachep); -#else -kmem_cache_destroy (cachep); -return 0 ; -#endif -} +#define kmem_cache_destroy (cachep) ({ kmem_cache_destroy(cachep); 0; })
Is this because it shall not even fail for older kernels?
Luis
PATCH 8/44 compat/2.6.20: add header's skeleton
In reply to Arnaud Lacombe--- include/linux/compat-2.6.20.h | 11 +++++++++++ 1 files changed, 11 insertions(+), 0 deletions(-) create mode 100644 include/linux/compat-2.6.20.h
include/linux/compat-2.6.20.h new file +11 -0
@@ -0,0 +1,11 @@ +#ifndef LINUX_26_20_COMPAT_H +#define LINUX_26_20_COMPAT_H + +#include <linux/version.h> + +/* Compat work for 2.6.20 */ +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,20 )) + +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20)) */ + +#endif /* LINUX_26_20_COMPAT_H */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 9/44 compat/2.6.20: add compat macro for `struct delayed_work'
In reply to Arnaud Lacombe--- include/linux/compat-2.6.20.h | 3 +++ 1 files changed, 3 insertions(+), 0 deletions(-)
include/linux/compat-2.6.20.h +3 -0
@@ -6,6 +6,9 @@/* Compat work for 2.6.20 */ #if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,20 )) +/* `struct work_struct' and `struct delayed_work' are field compatible */ +#define delayed_work work_struct +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,20)) */ #endif /* LINUX_26_20_COMPAT_H */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 10/44 compat/2.6.23: add compat inline for notifier_{from,to}_errno()
In reply to Arnaud Lacombe--- include/linux/compat-2.6.23.h | 14 ++++++++++++++ 1 files changed, 14 insertions(+), 0 deletions(-)
include/linux/compat-2.6.23.h +14 -0
@@ -7,11 +7,25 @@#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,23 ))#include <linux/netdevice.h> +#include <linux/notifier.h> #include <linux/sched.h> #include <linux/workqueue.h> #include <linux/genetlink.h> #include <net/sch_generic.h> +/* Encapsulate (negative) errno value (in particular, NOTIFY_BAD <=> EPERM). */ +static inline int notifier_from_errno (int err) +{ +return NOTIFY_STOP_MASK | (NOTIFY_OK - err); +} + +/* Restore (negative) errno value from notify return value. */ +static inline int notifier_to_errno (int ret) +{ + ret &= ~NOTIFY_STOP_MASK ; +return ret >NOTIFY_OK ?NOTIFY_OK - ret :0 ; +} +/* * Tell gcc if a function is cold. The compiler will assume any path * directly leading to the call is unlikely.
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 11/44 compat/2.6.24: add compat macro for sk_alloc()
In reply to Arnaud LacombeThis was part of the network namespace move.
--- include/linux/compat-2.6.24.h | 3 +++ 1 files changed, 3 insertions(+), 0 deletions(-)
include/linux/compat-2.6.24.h +3 -0
@@ -13,6 +13,7 @@#include <linux/types.h> #include <linux/list.h> #include <linux/scatterlist.h> +#include <net/sock.h> #define KEY_BLUETOOTH 237#define KEY_WLAN 238
@@ -161,6 +162,8 @@ struct ssb_device_id {/* Namespace stuff, introduced on 2.6.24 */ #define dev_get_by_index (a, b) dev_get_by_index(b)#define __dev_get_by_index (a, b) __dev_get_by_index(b) +#define sk_alloc (net, fam, pri, prot) sk_alloc(fam, pri, prot, 1) +/* * Display a 6 byte device address (MAC) in a readable format.
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 12/44 compat/2.6.25: don't include <linux/hw_random.h> prior to 2.6.18
In reply to Arnaud Lacombe--- include/linux/compat-2.6.25.h | 4 +++- 1 files changed, 3 insertions(+), 1 deletions(-)
include/linux/compat-2.6.25.h +3 -1
@@ -8,8 +8,10 @@#include <linux/types.h> #include <linux/io.h> -#include <linux/hw_random.h> #include <linux/leds.h> +#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,18 )) +#include <linux/hw_random.h> +#endif #include <linux/kernel.h> #include <linux/netdevice.h> #include <linux/pm.h>
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 13/44 compat/2.6.25: don't include <linux/{io,leds}.h> prior to 2.6.17
In reply to Arnaud Lacombe--- include/linux/compat-2.6.25.h | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/compat-2.6.25.h +2 -0
@@ -7,8 +7,10 @@#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,25 ))#include <linux/types.h> +#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,17 ))#include <linux/io.h> #include <linux/leds.h> +#endif #if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,18 ))#include <linux/hw_random.h> #endif
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 14/44 compat/2.6.25: conditionally defines `__WARN'
In reply to Arnaud Lacombepowerpc has a MD definition for it already.
--- include/linux/compat-2.6.25.h | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/compat-2.6.25.h +2 -0
@@ -72,7 +72,9 @@ int compat_pm_qos_power_deinit(void);/* Although pm_qos stuff is not implemented on <= 2.6.24 lets keep the define */ #define PM_QOS_DEFAULT_VALUE -1 +#ifndef __WARN#define __WARN (foo) dump_stack() +#endif #define dev_emerg(dev, format, arg...) \ dev_printk(KERN_EMERG , dev , format , ## arg)
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 15/44 compat/2.6.25: add compat macro for ip_route_output_key()
In reply to Arnaud LacombeThis is was a late network namespace conversion :)
--- include/linux/compat-2.6.25.h | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/compat-2.6.25.h +2 -0
@@ -88,6 +88,8 @@ extern void __dev_addr_unsync(struct dev_addr_list **to, int *to_count, struct#define seq_file_net &init_net; +#define ip_route_output_key (net, rt, fl) ip_route_output_key(rt, fl) +enum nf_inet_hooks {NF_INET_PRE_ROUTING =0 ,NF_INET_LOCAL_IN =1 ,
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 16/44 compat/2.6.29: don't include <linux/if_link.h> prior to 2.6.19
In reply to Arnaud Lacombe--- include/linux/compat-2.6.29.h | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/compat-2.6.29.h +2 -0
@@ -3,7 +3,9 @@#include <linux/version.h> #include <linux/netdevice.h> +#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,19 ))#include <linux/if_link.h> +#endif /* * I kow this looks odd.. but 2.6.32 added the netdev_tx_t
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 17/44 compat/2.6.29: const'ify compat inline to dev_get_stats()
In reply to Arnaud Lacombe--- include/linux/compat-2.6.29.h | 3 ++- 1 files changed, 2 insertions(+), 1 deletions(-)
include/linux/compat-2.6.29.h +2 -1
@@ -229,7 +229,8 @@ static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list, } -static inline struct net_device_stats *dev_get_stats (struct net_device *dev) +static inline const struct net_device_stats * +dev_get_stats (struct net_device *dev) {return dev->get_stats (dev); }
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 18/44 compat/2.6.31: gather all the threaded IRQ backport together
In reply to Arnaud Lacombe--- include/linux/compat-2.6.31.h | 19 +++++++++---------- 1 files changed, 9 insertions(+), 10 deletions(-)
include/linux/compat-2.6.31.h +9 -10
@@ -58,16 +58,6 @@#define SDIO_DEVICE_ID_MARVELL_8688WLAN 0x9104#endif -struct compat_threaded_irq { -unsigned int irq; -irq_handler_t handler; -irq_handler_t thread_fn; -void *dev_id; -char wq_name[64 ]; -struct workqueue_struct *wq; -struct work_struct work; -}; -/* * kmemleak was introduced on 2.6.31, since older kernels do not have * we simply ignore its tuning.
@@ -115,6 +105,15 @@ static inline struct rtable *skb_rtable(const struct sk_buff *skb) }/* Backport threaded IRQ support */ +struct compat_threaded_irq { +unsigned int irq; +irq_handler_t handler; +irq_handler_t thread_fn; +void *dev_id; +char wq_name[64 ]; +struct workqueue_struct *wq; +struct work_struct work; +};static inline void compat_irq_work (struct work_struct *work)
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 19/44 compat/2.6.31: fix threaded irq backport wrt. INIT_WORK() macro
In reply to Arnaud Lacombe--- include/linux/compat-2.6.31.h | 14 ++++++++++++++ 1 files changed, 14 insertions(+), 0 deletions(-)
include/linux/compat-2.6.31.h +14 -0
@@ -115,12 +115,22 @@ struct compat_threaded_irq {struct work_struct work; }; +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,20 )) +static inline +void compat_irq_work (void *arg) +{ +struct work_struct *work = arg; + struct compat_threaded_irq *comp = container_of(work, struct compat_threaded_irq, work); + comp->thread_fn (comp->irq, comp->dev_id); +} +#else static inline void compat_irq_work (struct work_struct *work) { struct compat_threaded_irq *comp = container_of(work, struct compat_threaded_irq, work); comp->thread_fn (comp->irq, comp->dev_id); } +#endif static inline irqreturn_t compat_irq_dispatcher (int irq,void *dev_id)
@@ -150,7 +160,11 @@ int compat_request_threaded_irq(struct compat_threaded_irq *comp, comp->handler = handler; comp->thread_fn = thread_fn; comp->dev_id = dev_id; +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,20 )) +INIT_WORK (&comp->work, compat_irq_work, &comp->work); +#else INIT_WORK (&comp->work, compat_irq_work); +#endif if (!comp->wq) { snprintf(comp->wq_name, sizeof(comp->wq_name),
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 20/44 compat/2.6.32: fix compat inline to flush_delayed_work()
In reply to Arnaud LacombeThis was needed as major API changes was made in 2.6.21.
--- include/linux/compat-2.6.32.h | 5 +++++ 1 files changed, 5 insertions(+), 0 deletions(-)
include/linux/compat-2.6.32.h +5 -0
@@ -20,6 +20,10 @@static inline void flush_delayed_work (struct delayed_work *dwork) { +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,21 )) +cancel_delayed_work (dwork); +flush_scheduled_work (); +#else if (del_timer_sync (&dwork->timer)) {/* * This is what would happen on 2.6.32 but since we don't have
@@ -34,6 +38,7 @@ static inline void flush_delayed_work(struct delayed_work *dwork) */ }flush_work (&dwork->work); +#endif }/*
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 21/44 compat/2.6.32: add compat macros for sysctl helpers
In reply to Arnaud Lacombe--- include/linux/compat-2.6.32.h | 24 ++++++++++++++++++++++++ 1 files changed, 24 insertions(+), 0 deletions(-)
include/linux/compat-2.6.32.h +24 -0
@@ -99,6 +99,30 @@ struct dev_pm_ops name = { \#define dev_to_sdio_func (d) container_of(d, struct sdio_func, dev) +#define proc_dostring (table, write, buff, lenp, ppos) \ + ({ proc_dostring(table, write, NULL, buff, lenp, ppos); }) + +#define proc_dointvec (table, write, buff, lenp, ppos) \ + ({ proc_dointvec(table, write, NULL, buff, lenp, ppos); }) + +#define proc_dointvec_minmax (table, write, buff, lenp, ppos) \ + ({ proc_dointvec_minmax(table, write, NULL, buff, lenp, ppos); }) + +#define proc_dointvec_jiffies (table, write, buff, lenp, ppos) \ + ({ proc_dointvec_jiffies(table, write, NULL, buff, lenp, ppos); }) + +#define proc_dointvec_userhz_jiffies (table, write, buff, lenp, ppos) \ + ({ proc_dointvec_userhz_jiffies(table, write, NULL, buff, lenp, ppos); }) + +#define proc_dointvec_ms_jiffies (table, write, buff, lenp, ppos) \ + ({ proc_dointvec_ms_jiffies(table, write, NULL, buff, lenp, ppos); }) + +#define proc_doulongvec_minmax (table, write, buff, lenp, ppos) \ + ({ proc_doulongvec_minmax(table, write, NULL, buff, lenp, ppos); }) + +#define proc_doulongvec_ms_jiffies_minmax (table, write, buff, lenp, ppos) \ + ({ proc_doulongvec_ms_jiffies_minmax(table, write, NULL, buff, lenp, ppos); }) +#define lockdep_assert_held (l) do { } while (0)#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,32)) */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 22/44 compat/2.6.33: add compat macro to BUILD_BUG_ON_NOT_POWER_OF_2()
In reply to Arnaud Lacombe--- include/linux/compat-2.6.33.h | 3 +++ 1 files changed, 3 insertions(+), 0 deletions(-)
include/linux/compat-2.6.33.h +3 -0
@@ -14,6 +14,9 @@#include <linux/kfifo.h> #include <linux/firmware.h> +#define BUILD_BUG_ON_NOT_POWER_OF_2 (n) \ + BUILD_BUG_ON((n) == 0 || (((n) & ((n) - 1)) != 0)) +#define release_firmware compat_release_firmware#define request_firmware compat_request_firmware#define request_firmware_nowait compat_request_firmware_nowait
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 23/44 compat/2.6.33: fix pcmcia include
In reply to Arnaud LacombeThis fix 'ppc6xx_defconfig' build.
--- include/linux/compat-2.6.33.h | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/compat-2.6.33.h +2 -0
@@ -7,8 +7,10 @@#include <linux/skbuff.h> #if defined(CONFIG_PCCARD ) || defined(CONFIG_PCCARD_MODULE ) +struct pcmcia_socket ;#include <pcmcia/cs_types.h> #include <pcmcia/cistpl.h> +#include <pcmcia/cs.h> #include <pcmcia/ds.h> #endif #include <linux/kfifo.h>
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 24/44 compat/2.6.33: don't export prototype for pccard_loop_tuple()
In reply to Arnaud LacombeMainline kernel does not defines any prototype for this function.
--- include/linux/compat-2.6.33.h | 7 ------- 1 files changed, 0 insertions(+), 7 deletions(-)
include/linux/compat-2.6.33.h +0 -7
@@ -87,13 +87,6 @@ int pcmcia_loop_tuple(struct pcmcia_device *p_dev, cisdata_t code,#endif /* CONFIG_PCMCIA */ -/* loop over CIS entries */ -int pccard_loop_tuple (struct pcmcia_socket *s,unsigned int function, -cisdata_t code,cisparse_t *parse,void *priv_data, -int (*loop_tuple) (tuple_t *tuple, -cisparse_t *parse, -void *priv_data)); -#endif /* CONFIG_PCCARD */ /* Backport for kfifo
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 25/44 compat/2.6.36: `va_list' needs <stdarg.h>
In reply to Arnaud LacombeThis fixes the following build error:
include/linux/compat-2.6.36.h:13: error: syntax error before "va_list" include/linux/compat-2.6.36.h:13: warning: no semicolon at end of struct or union
--- include/linux/compat-2.6.36.h | 1 + 1 files changed, 1 insertions(+), 0 deletions(-)
include/linux/compat-2.6.36.h +1 -0
@@ -12,6 +12,7 @@#define kparam_block_sysfs_write (a)#define kparam_unblock_sysfs_write (a) +#include <stdarg.h> struct va_format {const char *fmt;va_list *va;
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 26/44 compat: remove a couple of macro definition protection
In reply to Arnaud Lacombe--- include/linux/compat-2.6.31.h | 4 ---- include/linux/compat-2.6.32.h | 2 -- 2 files changed, 0 insertions(+), 6 deletions(-)
include/linux/compat-2.6.31.h +0 -4
@@ -50,13 +50,9 @@#endif #endif -#ifndef NETDEV_PRE_UP #define NETDEV_PRE_UP 0x000D -#endif -#ifndef SDIO_DEVICE_ID_MARVELL_8688WLAN #define SDIO_DEVICE_ID_MARVELL_8688WLAN 0x9104 -#endif /* * kmemleak was introduced on 2.6.31, since older kernels do not have
include/linux/compat-2.6.32.h +0 -2
@@ -59,7 +59,6 @@ static inline void flush_delayed_work(struct delayed_work *dwork)#define SET_NETDEV_DEVTYPE (netdev, type) -#ifdef __KERNEL__/* Driver transmit return codes */ enum netdev_tx {BACKPORT_NETDEV_TX_OK =NETDEV_TX_OK ,/* driver took care of packet */
@@ -67,7 +66,6 @@ enum netdev_tx { BACKPORT_NETDEV_TX_LOCKED = NETDEV_TX_LOCKED, /* driver tx lock was already taken */ };typedef enum netdev_tx netdev_tx_t ; -#endif /* __KERNEL__ */ /* * dev_pm_ops is only available on kernels >= 2.6.29, for
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 27/44 compat/2.6.33: remove macro not in mainline
In reply to Arnaud Lacombe--- include/linux/compat-2.6.23.h | 4 ---- 1 files changed, 0 insertions(+), 4 deletions(-)
include/linux/compat-2.6.23.h +0 -4
@@ -84,10 +84,6 @@ static inline void tcf_destroy_chain_compat(struct tcf_proto **fl)#define da_users dmi_users#define da_gusers dmi_gusers -/* dev_set_promiscuity() was moved to __dev_set_promiscuity() on 2.6.23 and -* dev_set_promiscuity() became a wrapper. */ -#define __dev_set_promiscuity dev_set_promiscuity -/* Our own 2.6.22 port on compat.c */ extern void dev_mc_unsync (struct net_device *to,struct net_device *from);extern int dev_mc_sync (struct net_device *to,struct net_device *from);
Signature · 1 line
1.7.2.30.gc37d7.dirty
Re: [PATCH 27/44] compat/2.6.33: remove macro not in mainline
Quoted 10 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>--- include/linux/compat-2.6.23.h | 4 ---- 1 files changed, 0 insertions(+), 4 deletions(-)include/linux/compat-2.6.23.h +0
-0
include/linux/compat-2.6.23.h static inline void tcf_destroy_chain_compat(struct tcf_proto **fl) · line 90
Quoted 7 lines
include/linux/compat-2.6.23.h +0
-4@@ -84,10 +84,6 @@ static inline void tcf_destroy_chain_compat(struct tcf_proto **fl)#define da_users dmi_users#define da_gusers dmi_gusers -/* dev_set_promiscuity() was moved to __dev_set_promiscuity() on 2.6.23 and -* dev_set_promiscuity() became a wrapper. */ -#define __dev_set_promiscuity dev_set_promiscuity -
Good call, how did you find these non-in-kernel thingies?
Luis
PATCH 28/44 compat: avoid using `#include_next' directive in compat headers
In reply to Arnaud Lacombe--- include/linux/pm_qos_params.h | 7 +++---- include/linux/tracepoint.h | 14 +++++++------- include/net/net_namespace.h | 14 +++++++------- include/trace/define_trace.h | 5 ----- 4 files changed, 17 insertions(+), 23 deletions(-)
include/linux/pm_qos_params.h +3 -4
@@ -1,8 +1,7 @@#include <linux/version.h> -#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,25 )) -#include_next <linux/pm_qos_params.h> -#else +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,25 )) +/* interface for the pm_qos_power infrastructure of the linux kernel. * * Mark Gross <[email protected]>
@@ -28,5 +27,5 @@ int pm_qos_requirement(int qos);int pm_qos_add_notifier (int qos,struct notifier_block *notifier);int pm_qos_remove_notifier (int qos,struct notifier_block *notifier); -#endif /* (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)) */ +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)) */
include/linux/tracepoint.h +7 -7
@@ -1,10 +1,10 @@ -#ifndef _COMPAT_LINUX_TRACEPOINT_H -#define _COMPAT_LINUX_TRACEPOINT_H 1 -#include <linux/version.h> -#if (LINUX_VERSION_CODE >KERNEL_VERSION (2 ,6 ,27 )) -#include_next <linux/tracepoint.h> -#endif /* (LINUX_VERSION_CODE > KERNEL_VERSION(2,6,27)) */ +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,28 )) + +#ifndef _LINUX_TRACEPOINT_H +#define _LINUX_TRACEPOINT_H + +#endif /* _LINUX_TRACEPOINT_H */ -#endif /* _COMPAT_LINUX_TRACEPOINT_H */ +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)) */
include/net/net_namespace.h +7 -7
@@ -1,10 +1,10 @@ -#ifndef _COMPAT_NET_NET_NAMESPACE_H -#define _COMPAT_NET_NET_NAMESPACE_H 1 -#include <linux/version.h> -#if (LINUX_VERSION_CODE >KERNEL_VERSION (2 ,6 ,23 )) -#include_next <net/net_namespace.h> -#endif /* (LINUX_VERSION_CODE > KERNEL_VERSION(2,6,23)) */ +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,24 )) + +#ifndef __NET_NET_NAMESPACE_H +#define __NET_NET_NAMESPACE_H + +#endif /* __NET_NET_NAMESPACE_H */ -#endif /* _COMPAT_NET_NET_NAMESPACE_H */ +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24)) */
include/trace/define_trace.h +0 -5
@@ -1,5 +0,0 @@ -#include <linux/version.h> - -#if (LINUX_VERSION_CODE >KERNEL_VERSION (2 ,6 ,30 )) -#include_next <trace/define_trace.h> -#endif /* (LINUX_VERSION_CODE > KERNEL_VERSION(2,6,30)) */
Signature · 1 line
1.7.2.30.gc37d7.dirty
Re: [PATCH 28/44] compat: avoid using `#include_next' directive in compat headers
Quoted 24 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>--- include/linux/pm_qos_params.h | 7 +++---- include/linux/tracepoint.h | 14 +++++++------- include/net/net_namespace.h | 14 +++++++------- include/trace/define_trace.h | 5 ----- 4 files changed, 17 insertions(+), 23 deletions(-)include/linux/pm_qos_params.h +2
-3@@ -1,8 +1,7 @@#include <linux/version.h> -#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,25 )) -#include_next <linux/pm_qos_params.h> -#else +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,25 )) +/* interface for the pm_qos_power infrastructure of the linux kernel. * * Mark Gross <[email protected]>
include/linux/pm_qos_params.h int pm_qos_requirement(int qos); · line 31
Quoted 6 lines
include/linux/pm_qos_params.h +1
-1@@ -28,5 +27,5 @@ int pm_qos_requirement(int qos);int pm_qos_add_notifier (int qos,struct notifier_block *notifier);int pm_qos_remove_notifier (int qos,struct notifier_block *notifier); -#endif /* (LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)) */ +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,25)) */
With #include_next gone, how would the compiler find the other header file for version >= 2.6.25?
- Felix
Re: [PATCH 28/44] compat: avoid using `#include_next' directive in compat headers
In reply to Arnaud LacombeQuoted 13 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>--- include/linux/pm_qos_params.h | 7 +++---- include/linux/tracepoint.h | 14 +++++++------- include/net/net_namespace.h | 14 +++++++------- include/trace/define_trace.h | 5 ----- 4 files changed, 17 insertions(+), 23 deletions(-)include/linux/pm_qos_params.h +0
-0
include/linux/pm_qos_params.h line 4
Quoted 7 lines
include/linux/pm_qos_params.h +2
-3@@ -1,8 +1,7 @@#include <linux/version.h> -#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,25 )) -#include_next <linux/pm_qos_params.h> -#else +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,25 )) +
But will the kernels' own pm_qos_params.h be included for kernels >= 2.6.25?
Luis
Re: [PATCH 28/44] compat: avoid using `#include_next' directive in compat headers
Hi,
Quoted 16 lines · Luis R. Rodriguez
Quoted 15 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>---
include/linux/pm_qos_params.h | 7 +++----include/linux/tracepoint.h| 14 +++++++-------include/net/net_namespace.h| 14 +++++++-------include/trace/define_trace.h| 5 -----4 files changed, 17 insertions(+), 23 deletions(-)include/linux/pm_qos_params.h +0
-0@@ -1,8 +1,7 @@ #include <linux/version.h>
Quoted 8 lines
include/linux/pm_qos_params.h line 4
Quoted 5 lines
include/linux/pm_qos_params.h +2
-3-#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,25 )) -#include_next <linux/pm_qos_params.h> -#else +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,25 )) +But will the kernels' own
pm_qos_params.hbe included for kernels >= 2.6.25?
The compat headers should be at the end of the include list, so that the kernel headers get included first, and the compat one will only be when the kernel does not provide the header. This is the only sane way to override kernel provided stuff. That said, there is certainly a use-case I missed.
- Arnaud
Re: [PATCH 28/44] compat: avoid using `#include_next' directive in compat headers
Quoted 28 lines · Arnaud Lacombe
Hi,
Quoted 16 lines · Luis R. Rodriguez
Quoted 15 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>---
include/linux/pm_qos_params.h | 7 +++----include/linux/tracepoint.h| 14 +++++++-------include/net/net_namespace.h| 14 +++++++-------include/trace/define_trace.h| 5 -----4 files changed, 17 insertions(+), 23 deletions(-)include/linux/pm_qos_params.h +0
-0@@ -1,8 +1,7 @@ #include <linux/version.h>Quoted 8 lines
include/linux/pm_qos_params.h line 4
Quoted 5 lines
include/linux/pm_qos_params.h +2
-3-#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,25 )) -#include_next <linux/pm_qos_params.h> -#else +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,25 )) +But will the kernels' own
pm_qos_params.hbe included for kernels >= 2.6.25?
Quoted 5 lines
The compat headers should be at the end of the include list, so that the kernel headers get included first, and the compat one will only be when the kernel does not provide the header. This is the only sane way to override kernel provided stuff. That said, there is certainly a use-case I missed.
Agreed, but you are missing the purpose of the trick used here. The purpose of the include_next was so that we can name our own <linux/pm_qos_params.h> which is part of compat and these directories *will* get a priority over the kernel's so that way we can avoid ifdef'ing all includes for the same file on the upstream code. Without an include_next we'd simply skip including the kernel's own original header file.
Luis
Re: [PATCH 28/44] compat: avoid using `#include_next' directive in compat headers
Hi,
Quoted 9 lines · Luis R. Rodriguez
Quoted 6 lines · Arnaud Lacombe
[...]
The compat headers should be at the end of the include list, so that the kernel headers get included first, and the compat one will only be when the kernel does not provide the header. This is the only sane way to override kernel provided stuff. That said, there is certainly a use-case I missed.Agreed, but you are missing the purpose of the trick used here.
certainly :)
Quoted 5 lines
The
purpose of theinclude_nextwas so that we can name our own <linux/pm_qos_params.h> which is part of compat and these directories *will* get a priority over the kernel's so that way we can avoid ifdef'ing all includes for the same file on the upstream code.
I'm not sure to get what you mean by "we can avoid ifdef'ing all includes for the same file on the upstream code", can you details ?
Quoted 4 lines
Without
aninclude_nextwe'd simply skip including the kernel's own original header file.
I admit that I removed that part merely because I went loose on the trigger that day. AFAIR, there was no technical point per-se, so I would not mind omitting this patch.
- Arnaud
Re: [PATCH 28/44] compat: avoid using `#include_next' directive in compat headers
Quoted 15 lines · Arnaud Lacombe
Hi,
Quoted 9 lines · Luis R. Rodriguez
Quoted 6 lines · Arnaud Lacombe
[...]
The compat headers should be at the end of the include list, so that the kernel headers get included first, and the compat one will only be when the kernel does not provide the header. This is the only sane way to override kernel provided stuff. That said, there is certainly a use-case I missed.Agreed, but you are missing the purpose of the trick used here.
certainly :)
Quoted 7 lines
Quoted 5 lines
The
purpose of theinclude_nextwas so that we can name our own <linux/pm_qos_params.h> which is part of compat and these directories *will* get a priority over the kernel's so that way we can avoid ifdef'ing all includes for the same file on the upstream code.I'm not sure to get what you mean by "we can avoid ifdef'ing all includes for the same file on the upstream code", can you details ?
Ah, that's right, you don't see or use compat-wireless likely. Ok, so what we do for backporting the 802.11 subsystem, Bluetooth and some Ethernet drivers is we take the stock upstream files, stuff them into a new directory, and then apply a series of patches to them to get the files to properly compile. The patches address all things that was not possible to backport through compat.git. Header file changes is one of the typical things we run into. If a new header file is added to newer kernels we can play a hack for older kernels and leave the code intact and introduce our own include/linux/foo.h through compat by prefering the local path over the kernel's path. If we're on newer kernels though include_next will go on the search path and find the kernel's own header file.
Quoted 7 lines
Quoted 4 lines
Without
aninclude_nextwe'd simply skip including the kernel's own original header file.I admit that I removed that part merely because I went loose on the trigger that day. AFAIR, there was no technical point per-se, so I would not mind omitting this patch.
OK cool, lets skip that then.
Luis
Re: [PATCH 28/44] compat: avoid using `#include_next' directive in compat headers
Hi,
Quoted 24 lines · Luis R. Rodriguez
Quoted 15 lines · Arnaud Lacombe
Hi,
Quoted 9 lines · Luis R. Rodriguez
Quoted 6 lines · Arnaud Lacombe
[...]
The compat headers should be at the end of the include list, so that the kernel headers get included first, and the compat one will only be when the kernel does not provide the header. This is the only sane way to override kernel provided stuff. That said, there is certainly a use-case I missed.Agreed, but you are missing the purpose of the trick used here.
certainly :)
Quoted 7 lines
Quoted 5 lines
The
purpose of theinclude_nextwas so that we can name our own <linux/pm_qos_params.h> which is part of compat and these directories *will* get a priority over the kernel's so that way we can avoid ifdef'ing all includes for the same file on the upstream code.I'm not sure to get what you mean by "we can avoid ifdef'ing all includes for the same file on the upstream code", can you details ?
Quoted 13 lines
Ah, that's right, you don't see or use compat-wireless likely. Ok, so what we do for backporting the 802.11 subsystem, Bluetooth and some Ethernet drivers is we take the stock upstream files, stuff them into a new directory, and then apply a series of patches to them to get the files to properly compile. The patches address all things that was not possible to backport through compat.git. Header file changes is one of the typical things we run into. If a new header file is added to newer kernels we can play a hack for older kernels and leave the code intact and introduce our own
include/linux/foo.hthrough compat by prefering the local path over the kernel's path. If we're on newer kernels thoughinclude_nextwill go on the search path and find the kernel's own header file.
Is there any reason you want the compat's include directory to be *before* the kernel include directory ? Because just considering this issue, if the kernel include directory was seen first, this would not arise, as the compat-provided header would just be a fallback.
- Arnaud
Re: [PATCH 28/44] compat: avoid using `#include_next' directive in compat headers
Quoted 41 lines · Arnaud Lacombe
Hi,
Quoted 24 lines · Luis R. Rodriguez
Quoted 15 lines · Arnaud Lacombe
Hi,
Quoted 9 lines · Luis R. Rodriguez
Quoted 6 lines · Arnaud Lacombe
[...]
The compat headers should be at the end of the include list, so that the kernel headers get included first, and the compat one will only be when the kernel does not provide the header. This is the only sane way to override kernel provided stuff. That said, there is certainly a use-case I missed.Agreed, but you are missing the purpose of the trick used here.
certainly :)
Quoted 7 lines
Quoted 5 lines
The
purpose of theinclude_nextwas so that we can name our own <linux/pm_qos_params.h> which is part of compat and these directories *will* get a priority over the kernel's so that way we can avoid ifdef'ing all includes for the same file on the upstream code.I'm not sure to get what you mean by "we can avoid ifdef'ing all includes for the same file on the upstream code", can you details ?
Quoted 13 lines
Ah, that's right, you don't see or use compat-wireless likely. Ok, so what we do for backporting the 802.11 subsystem, Bluetooth and some Ethernet drivers is we take the stock upstream files, stuff them into a new directory, and then apply a series of patches to them to get the files to properly compile. The patches address all things that was not possible to backport through compat.git. Header file changes is one of the typical things we run into. If a new header file is added to newer kernels we can play a hack for older kernels and leave the code intact and introduce our own
include/linux/foo.hthrough compat by prefering the local path over the kernel's path. If we're on newer kernels thoughinclude_nextwill go on the search path and find the kernel's own header file.
Quoted 2 lines
Is there any reason you want the compat's include directory to be *before* the kernel include directory ?
Hm, yeah but I forget why. It'll come to me.
Quoted 3 lines
Because just considering this
issue, if the kernel include directory was seen first, this would not arise, as the compat-provided header would just be a fallback.
True.
Luis
Re: [PATCH 28/44] compat: avoid using `#include_next' directive in compat headers
Quoted 42 lines · Luis R. Rodriguez
Quoted 41 lines · Arnaud Lacombe
Hi,
Quoted 37 lines · Luis R. Rodriguez
Quoted 15 lines · Arnaud Lacombe
Hi,
Quoted 9 lines · Luis R. Rodriguez
Quoted 6 lines · Arnaud Lacombe
[...]
The compat headers should be at the end of the include list, so that the kernel headers get included first, and the compat one will only be when the kernel does not provide the header. This is the only sane way to override kernel provided stuff. That said, there is certainly a use-case I missed.Agreed, but you are missing the purpose of the trick used here.
certainly :)
Quoted 7 lines
Quoted 5 lines
The
purpose of theinclude_nextwas so that we can name our own <linux/pm_qos_params.h> which is part of compat and these directories *will* get a priority over the kernel's so that way we can avoid ifdef'ing all includes for the same file on the upstream code.I'm not sure to get what you mean by "we can avoid ifdef'ing all includes for the same file on the upstream code", can you details ?
Ah, that's right, you don't see or use compat-wireless likely. Ok, so what we do for backporting the 802.11 subsystem, Bluetooth and some Ethernet drivers is we take the stock upstream files, stuff them into a new directory, and then apply a series of patches to them to get the files to properly compile. The patches address all things that was not possible to backport through compat.git. Header file changes is one of the typical things we run into. If a new header file is added to newer kernels we can play a hack for older kernels and leave the code intact and introduce our own
include/linux/foo.hthrough compat by prefering the local path over the kernel's path. If we're on newer kernels thoughinclude_nextwill go on the search path and find the kernel's own header file.
Quoted 4 lines
Quoted 2 lines
Is there any reason you want the compat's include directory to be *before* the kernel include directory ?
Hm, yeah but I forget why. It'll come to me.
Ah, OK so its because we provide our own mac80211.h, and new cfg80211.h
Luis
PATCH 29/44 compat/2.6.22: rework `uninitialized_var' and `__maybe_unused' macros
In reply to Arnaud LacombeProtect their definitions, and add fallbacks.
--- include/linux/compat-2.6.22.h | 13 +++++++++++-- 1 files changed, 11 insertions(+), 2 deletions(-)
include/linux/compat-2.6.22.h +11 -2
@@ -80,9 +80,18 @@ static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,memcpy (to, skb->data + offset, len); } -#define __maybe_unused __attribute__((unused)) +#if defined(__GNUC__) +#define __maybe_unused __attribute__((unused)) +#define uninitialized_var (x) x = x +#endif + +#ifndef __maybe_unused +#define __maybe_unused/* unimplemented */ +#endif -#define uninitialized_var (x) x = x +#ifndef uninitialized_var +#define uninitialized_var (x) x +#endif /* This will lead to very weird behaviour... */ #define NLA_BINARY NLA_STRING
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 30/44 compat/2.6.22: nuke pci_{clear,set}_mwi()
In reply to Arnaud LacombeAll 2.6 kernel have had them.
--- include/linux/compat-2.6.22.h | 9 --------- 1 files changed, 0 insertions(+), 9 deletions(-)
include/linux/compat-2.6.22.h +0 -9
@@ -96,15 +96,6 @@ static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,/* This will lead to very weird behaviour... */ #define NLA_BINARY NLA_STRING -static inline int pci_set_mwi (struct pci_dev *dev) -{ -return -ENOSYS ; -} - -static inline void pci_clear_mwi (struct pci_dev *dev) -{ -} -#define list_first_entry (ptr, type, member) \ list_entry((ptr)->next, type, member)
Signature · 1 line
1.7.2.30.gc37d7.dirty
Re: [PATCH 30/44] compat/2.6.22: nuke pci_{clear,set}_mwi()
Quoted 1 line · Arnaud Lacombe
All 2.6 kernel have had them.
Odd, wonder why we added it. Ah well, away it goes! Thanks.
Luis
PATCH 31/44 compat/2.6.22: add compat for couple of TCP accessors
In reply to Arnaud Lacombe--- include/linux/compat-2.6.22.h | 20 ++++++++++++++++++++ 1 files changed, 20 insertions(+), 0 deletions(-)
include/linux/compat-2.6.22.h +20 -0
@@ -80,6 +80,26 @@ static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,memcpy (to, skb->data + offset, len); } +static inline struct iphdr *ip_hdr (const struct sk_buff *skb) +{ +return (struct iphdr *)skb_network_header (skb); +} + +static inline struct tcphdr *tcp_hdr (const struct sk_buff *skb) +{ +return (struct tcphdr *)skb_transport_header (skb); +} + +static inline unsigned int tcp_hdrlen (const struct sk_buff *skb) +{ +return tcp_hdr (skb)->doff *4 ; +} + +static inline unsigned int tcp_optlen (const struct sk_buff *skb) +{ +return (tcp_hdr (skb)->doff -5 ) *4 ; +} +#if defined(__GNUC__)#define __maybe_unused __attribute__((unused))#define uninitialized_var (x) x = x
Signature · 1 line
1.7.2.30.gc37d7.dirty
Re: [PATCH 31/44] compat/2.6.22: add compat for couple of TCP accessors
Hi,
Quoted 10 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>---
include/linux/compat-2.6.22.h | 20 ++++++++++++++++++++1 files changed, 20 insertions(+), 0 deletions(-)include/linux/compat-2.6.22.h +0
-0
include/linux/compat-2.6.22.h static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, · line 87
Quoted 8 lines
include/linux/compat-2.6.22.h +5
-0@@ -80,6 +80,26 @@ static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb, memcpy(to, skb->data + offset, len); } +static inline struct iphdr *ip_hdr (const struct sk_buff *skb) +{ + return (struct iphdr *)skb_network_header(skb); +} +
Actually, this one is already defined a few lines above, but lost in the middle of skb accessor inlines :/ I'll remove this part.
- Arnaud
PATCH 32/44 compat/2.6.22: add compat inline for ip_hdrlen()
In reply to Arnaud Lacombe--- include/linux/compat-2.6.22.h | 5 +++++ 1 files changed, 5 insertions(+), 0 deletions(-)
include/linux/compat-2.6.22.h +5 -0
@@ -85,6 +85,11 @@ static inline struct iphdr *ip_hdr(const struct sk_buff *skb)return (struct iphdr *)skb_network_header (skb); } +static inline unsigned int ip_hdrlen (const struct sk_buff *skb) +{ +return ip_hdr (skb)->ihl *4 ; +} +static inline struct tcphdr *tcp_hdr (const struct sk_buff *skb) {return (struct tcphdr *)skb_transport_header (skb);
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 33/44 compat/2.6.22: constify skb_transport_header()'s argument
In reply to Arnaud Lacombe--- include/linux/compat-2.6.22.h | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-)
include/linux/compat-2.6.22.h +1 -1
@@ -46,7 +46,7 @@ static inline void skb_set_transport_header(struct sk_buff *skb, int offset) skb->h.raw = skb->data + offset; } -static inline unsigned char *skb_transport_header (struct sk_buff *skb) +static inline unsigned char *skb_transport_header (const struct sk_buff *skb) {return skb->h.raw; }
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 34/44 compat/2.6.24: move 'struct net' to its own compat header
In reply to Arnaud Lacombe--- include/linux/compat-2.6.24.h | 31 ------------------------------- include/net/net_namespace.h | 36 ++++++++++++++++++++++++++++++++++++ 2 files changed, 36 insertions(+), 31 deletions(-)
include/linux/compat-2.6.24.h +0 -31
@@ -21,37 +21,6 @@#define DMA_BIT_MASK (n) (((n) == 64) ? ~0ULL : ((1ULL<<(n))-1)) -struct proc_dir_entry ; -struct net_device ; -struct net { -atomic_t count;/* To decided when the network -* namespace should be freed. -*/ -atomic_t use_count;/* To track references we -* destroy on demand -*/ -struct list_head list;/* list of network namespaces */ -struct work_struct work;/* work struct for freeing */ - -struct proc_dir_entry *proc_net; -struct proc_dir_entry *proc_net_stat; -struct proc_dir_entry *proc_net_root; - -struct net_device *loopback_dev;/* The loopback */ - -struct list_head dev_base_head; -struct hlist_head *dev_name_head; -struct hlist_head *dev_index_head; -}; - -#ifdef CONFIG_NET -/* Init's network namespace */ -extern struct net init_net; -#define INIT_NET_NS (net_ns) .net_ns = &init_net, -#else -#define INIT_NET_NS (net_ns) -#endif -/* Added on 2.6.24 in include/linux/types.h by Al viro on commit 142956af */ typedef unsigned long uintptr_t ;
include/net/net_namespace.h +36 -0
@@ -5,6 +5,42 @@#ifndef __NET_NET_NAMESPACE_H#define __NET_NET_NAMESPACE_H +#include <linux/list.h> +#include <linux/workqueue.h> + +#include <asm/atomic.h> + +struct proc_dir_entry ; +struct net_device ; +struct net { +atomic_t count;/* To decided when the network +* namespace should be freed. +*/ +atomic_t use_count;/* To track references we +* destroy on demand +*/ +struct list_head list;/* list of network namespaces */ +struct work_struct work;/* work struct for freeing */ + +struct proc_dir_entry *proc_net; +struct proc_dir_entry *proc_net_stat; +struct proc_dir_entry *proc_net_root; + +struct net_device *loopback_dev;/* The loopback */ + +struct list_head dev_base_head; +struct hlist_head *dev_name_head; +struct hlist_head *dev_index_head; +}; + +#ifdef CONFIG_NET +/* Init's network namespace */ +extern struct net init_net; +#define INIT_NET_NS (net_ns) .net_ns = &init_net, +#else +#define INIT_NET_NS (net_ns) +#endif +#endif /* __NET_NET_NAMESPACE_H */ #endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,24)) */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 35/44 compat/2.6.26: {U,}SHORT_MAX/SHORT_MIN have been deprecated
In reply to Arnaud Lacombe--- include/linux/compat-2.6.26.h | 5 ----- 1 files changed, 0 insertions(+), 5 deletions(-)
include/linux/compat-2.6.26.h +0 -5
@@ -39,11 +39,6 @@#define SW_RFKILL_ALL 0x03/* rfkill master switch, type "any" set = radio enabled */ -/* From kernel.h */ -#define USHORT_MAX ((u16)(~0U)) -#define SHORT_MAX ((s16)(USHORT_MAX>>1)) -#define SHORT_MIN (-SHORT_MAX - 1) -extern int dev_set_name (struct device *dev,const char *name, ...) __attribute__((format (printf,2 ,3 )));
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 36/44 compat/2.6.26: move unaligned backport to its own header
In reply to Arnaud Lacombe--- include/linux/asm/unaligned.h | 194 +++++++++++++++++++++++++++++++++++++ include/linux/compat-2.6.26.h | 214 +---------------------------------------- 2 files changed, 197 insertions(+), 211 deletions(-) create mode 100644 include/linux/asm/unaligned.h
include/linux/asm/unaligned.h new file +194 -0
@@ -0,0 +1,194 @@ +#ifndef _COMPAT_ASM_UNALIGNED_H +#define _COMPAT_ASM_UNALIGNED_H + +#include <linux/version.h> + +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,26 )) + +/* +* 2.6.26 added its own unaligned API which the +* new drivers can use. Lets port it here by including it in older +* kernels and also deal with the architecture handling here. +*/ + +#ifdef CONFIG_ALPHA +#include <linux/unaligned/be_struct.h> +#include <linux/unaligned/le_byteshift.h> +#include <linux/unaligned/generic.h> +#endif /* alpha */ + +#ifdef CONFIG_ARM +#include <linux/unaligned/le_byteshift.h> +#include <linux/unaligned/be_byteshift.h> +#include <linux/unaligned/generic.h> +#endif /* arm */ + +#ifdef CONFIG_AVR32 +/* +* AVR32 can handle some unaligned accesses, depending on the +* implementation. The AVR32 AP implementation can handle unaligned +* words, but halfwords must be halfword-aligned, and doublewords must +* be word-aligned. +* +* However, swapped word loads must be word-aligned so we can't +* optimize word loads in general. +*/ +#include <linux/unaligned/be_struct.h> +#include <linux/unaligned/le_byteshift.h> +#include <linux/unaligned/generic.h> +#endif + +#ifdef CONFIG_BLACKFIN +#include <linux/unaligned/le_struct.h> +#include <linux/unaligned/be_byteshift.h> +#include <linux/unaligned/generic.h> +#endif /* blackfin */ + +#ifdef CONFIG_CRIS +/* +* CRIS can do unaligned accesses itself. +*/ +#include <linux/unaligned/access_ok.h> +#include <linux/unaligned/generic.h> +#endif /* cris */ + +#ifdef CONFIG_FRV +#include <linux/unaligned/le_byteshift.h> +#include <linux/unaligned/be_byteshift.h> +#include <linux/unaligned/generic.h> +#endif /* frv */ + +#ifdef CONFIG_H8300 +#include <linux/unaligned/be_memmove.h> +#include <linux/unaligned/le_byteshift.h> +#include <linux/unaligned/generic.h> +#endif /* h8300 */ + +#ifdef CONFIG_IA64 +#include <linux/unaligned/le_struct.h> +#include <linux/unaligned/be_byteshift.h> +#include <linux/unaligned/generic.h> +#endif /* ia64 */ + +#ifdef CONFIG_M32R +#if defined(__LITTLE_ENDIAN__) +# include <linux/unaligned/le_memmove.h> +# include <linux/unaligned/be_byteshift.h> +# include <linux/unaligned/generic.h> +#else +# include <linux/unaligned/be_memmove.h> +# include <linux/unaligned/le_byteshift.h> +# include <linux/unaligned/generic.h> +#endif +#endif /* m32r */ + +#ifdef CONFIG_M68K /* this handles both m68k and m68knommu */ +#ifdef CONFIG_COLDFIRE +#include <linux/unaligned/be_struct.h> +#include <linux/unaligned/le_byteshift.h> +#include <linux/unaligned/generic.h> +#else +/* +* The m68k can do unaligned accesses itself. +*/ +#include <linux/unaligned/access_ok.h> +#include <linux/unaligned/generic.h> +#endif +#endif /* m68k and m68knommu */ + +#ifdef CONFIG_MIPS +#if defined(__MIPSEB__) +# include <linux/unaligned/be_struct.h> +# include <linux/unaligned/le_byteshift.h> +# include <linux/unaligned/generic.h> +# define get_unaligned __get_unaligned_be +# define put_unaligned __put_unaligned_be +#elif defined(__MIPSEL__) +# include <linux/unaligned/le_struct.h> +# include <linux/unaligned/be_byteshift.h> +# include <linux/unaligned/generic.h> +#endif +#endif /* mips */ + +#ifdef CONFIG_MN10300 +#include <linux/unaligned/access_ok.h> +#include <linux/unaligned/generic.h> +#endif /* mn10300 */ + +#ifdef CONFIG_PARISC +#include <linux/unaligned/be_struct.h> +#include <linux/unaligned/le_byteshift.h> +#include <linux/unaligned/generic.h> +#endif /* parisc */ + +#ifdef CONFIG_PPC +/* +* The PowerPC can do unaligned accesses itself in big endian mode. +*/ +#include <linux/unaligned/access_ok.h> +#include <linux/unaligned/generic.h> +#endif /* ppc */ + +#ifdef CONFIG_S390 +/* +* The S390 can do unaligned accesses itself. +*/ +#include <linux/unaligned/access_ok.h> +#include <linux/unaligned/generic.h> +#endif /* s390 */ + +#ifdef CONFIG_SUPERH +/* SH can't handle unaligned accesses. */ +#ifdef __LITTLE_ENDIAN__ +# include <linux/unaligned/le_struct.h> +# include <linux/unaligned/be_byteshift.h> +# include <linux/unaligned/generic.h> +#else +# include <linux/unaligned/be_struct.h> +# include <linux/unaligned/le_byteshift.h> +# include <linux/unaligned/generic.h> +#endif +#endif /* sh - SUPERH */ + +#ifdef CONFIG_SPARC +/* sparc and sparc64 */ +#include <linux/unaligned/be_struct.h> +#include <linux/unaligned/le_byteshift.h> +#include <linux/unaligned/generic.h> +#endif /* sparc */ + +#ifdef CONFIG_UML +#include "asm/arch/unaligned.h" +#endif /* um - uml */ + +#ifdef CONFIG_V850 +#include <linux/unaligned/be_byteshift.h> +#include <linux/unaligned/le_byteshift.h> +#include <linux/unaligned/generic.h> +#endif /* v850 */ + +#ifdef CONFIG_X86 +/* +* The x86 can do unaligned accesses itself. +*/ +#include <linux/unaligned/access_ok.h> +#include <linux/unaligned/generic.h> +#endif /* x86 */ + +#ifdef CONFIG_XTENSA +#ifdef __XTENSA_EL__ +# include <linux/unaligned/le_memmove.h> +# include <linux/unaligned/be_byteshift.h> +# include <linux/unaligned/generic.h> +#elif defined(__XTENSA_EB__) +# include <linux/unaligned/be_memmove.h> +# include <linux/unaligned/le_byteshift.h> +# include <linux/unaligned/generic.h> +#else +# error processor byte order undefined! +#endif +#endif /* xtensa */ + +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,26)) */ + +#endif /* _COMPAT_ASM_UNALIGNED_H */
include/linux/compat-2.6.26.h +3 -211
220 lines not on this page · raw message
Signature · 1 line
1.7.2.30.gc37d7.dirty
Re: [PATCH 36/44] compat/2.6.26: move unaligned backport to its own header
Quoted 6 lines · Arnaud Lacombe
Signed-off-by Arnaud Lacombe <[email protected]>--- include/linux/asm/unaligned.h | 194 +++++++++++++++++++++++++++++++++++++ include/linux/compat-2.6.26.h | 214 +---------------------------------------- 2 files changed, 197 insertions(+), 211 deletions(-) create mode 100644 include/linux/asm/unaligned.h
Good idea :)
Luis
PATCH 37/44 compat/2.6.2{7,8}: fix definition for IEEE80211_HT_CAP_DSSSCCK40
In reply to Arnaud Lacombe--- include/linux/compat-2.6.27.h | 3 --- include/linux/compat-2.6.28.h | 3 +++ 2 files changed, 3 insertions(+), 3 deletions(-)
include/linux/compat-2.6.27.h +0 -3
@@ -86,9 +86,6 @@ bool pci_pme_capable(struct pci_dev *dev, pci_power_t state);#define dma_mapping_error (pdev, dma_addr) dma_mapping_error(dma_addr)#define pci_dma_mapping_error (pdev, dma_addr) dma_mapping_error(pdev, dma_addr) -/* This is from include/linux/ieee80211.h */ -#define IEEE80211_HT_CAP_DSSSCCK40 0x1000 -/* New link list changes added as of 2.6.27, needed for ath9k */ static inline void __list_cut_position(struct list_head *list,
include/linux/compat-2.6.28.h +3 -0
@@ -234,6 +234,9 @@ extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,extern int n_tty_ioctl_helper (struct tty_struct *tty,struct file *file,unsigned int cmd,unsigned long arg); +/* This is from include/linux/ieee80211.h */ +#define IEEE80211_HT_CAP_DSSSCCK40 0x1000 +#endif /* (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,28)) */ #endif /* LINUX_26_28_COMPAT_H */
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 38/44 compat/2.6.2{8,9}: fix tracepoint macros
In reply to Arnaud LacombeIt was introduced in 2.6.29, not 2.6.28.
--- include/linux/compat-2.6.28.h | 24 ------------------------ include/linux/compat-2.6.29.h | 20 ++++++++++++++++++++ 2 files changed, 20 insertions(+), 24 deletions(-)
include/linux/compat-2.6.28.h +0 -24
@@ -192,30 +192,6 @@ static inline void skb_queue_splice_tail(const struct sk_buff_head *list, head->qlen += list->qlen; } } - -#ifndef DECLARE_TRACE - -#define TP_PROTO(args...) args -#define TP_ARGS(args...) args - -#define DECLARE_TRACE (name, proto, args) \ - static inline void _do_trace_##name(struct tracepoint *tp, proto) \ - { } \ - static inline void trace_##name(proto) \ - { } \ - static inline int register_trace_##name(void (*probe)(proto)) \ - { \ - return -ENOSYS; \ - } \ - static inline int unregister_trace_##name(void (*probe)(proto)) \ - { \ - return -ENOSYS; \ - } - -#define EXPORT_TRACEPOINT_SYMBOL_GPL (name) -#define EXPORT_TRACEPOINT_SYMBOL (name) - -#endif /* openSuse includes round_jiffies_up in it's kernel 2.6.27.
include/linux/compat-2.6.29.h +20 -0
@@ -277,6 +277,26 @@ static inline struct net *read_pnet(struct net * const *pnet)#endif +#define TP_PROTO(args...) args +#define TP_ARGS(args...) args + +#define DECLARE_TRACE (name, proto, args) \ + static inline void _do_trace_##name(struct tracepoint *tp, proto) \ + { } \ + static inline void trace_##name(proto) \ + { } \ + static inline int register_trace_##name(void (*probe)(proto)) \ + { \ + return -ENOSYS; \ + } \ + static inline int unregister_trace_##name(void (*probe)(proto)) \ + { \ + return -ENOSYS; \ + } + +#define EXPORT_TRACEPOINT_SYMBOL_GPL (name) +#define EXPORT_TRACEPOINT_SYMBOL (name) +extern int init_dummy_netdev (struct net_device *dev);#else
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 39/44 compat/2.6.29: protect headers inclusion
In reply to Arnaud Lacombe--- include/linux/compat-2.6.29.h | 5 +++-- 1 files changed, 3 insertions(+), 2 deletions(-)
include/linux/compat-2.6.29.h +3 -2
@@ -2,7 +2,10 @@#define LINUX_26_29_COMPAT_H #include <linux/version.h> +#include <linux/netdevice.h> + +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,29 ))#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,19 ))#include <linux/if_link.h> #endif
@@ -14,8 +17,6 @@*/ #include <linux/compat-2.6.32.h> -#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,29 )) -#include <linux/skbuff.h> #include <linux/usb.h>
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 40/44 compat/2.6.29: don't waste space around DIV_ROUND_CLOSEST()
In reply to Arnaud Lacombe--- include/linux/compat-2.6.29.h | 6 ++---- 1 files changed, 2 insertions(+), 4 deletions(-)
include/linux/compat-2.6.29.h +2 -4
@@ -242,12 +242,10 @@ extern void usb_unpoison_anchored_urbs(struct usb_anchor *anchor);#endif /* CONFIG_USB */ #endif -#define DIV_ROUND_CLOSEST (x, divisor)( \ -{ \ +#define DIV_ROUND_CLOSEST (x, divisor) ({ \ typeof(divisor) __divisor = divisor; \ (((x) + ((__divisor) / 2)) / (__divisor)); \ -} \ -) +})extern int eth_mac_addr (struct net_device *dev,void *p);extern int eth_change_mtu (struct net_device *dev,int new_mtu);
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 41/44 compat/2.6.29: fix typos
In reply to Arnaud Lacombe--- include/linux/compat-2.6.29.h | 4 ++-- 1 files changed, 2 insertions(+), 2 deletions(-)
include/linux/compat-2.6.29.h +2 -2
@@ -11,8 +11,8 @@#endif /* -* I kow this looks odd.. but 2.6.32 added the netdev_tx_t -* and we backport that there so inlcude that header first +* I know this looks odd.. but 2.6.32 added the netdev_tx_t +* and we backport that there so include that header first * as we need it for the netdev ops. */ #include <linux/compat-2.6.32.h>
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 42/44 compat/2.6.{21,30}: make dev_set_uevent_suppress() a no-op for <2.6.21
In reply to Arnaud Lacombe--- include/linux/compat-2.6.30.h | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)
include/linux/compat-2.6.30.h +2 -0
@@ -27,7 +27,9 @@ enum dpm_order {static inline void dev_set_uevent_suppress (struct device *dev,int val) { +#if (LINUX_VERSION_CODE >=KERNEL_VERSION (2 ,6 ,21 )) dev->uevent_suppress = val; +#endif }/*
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 43/44 compat/2.6.{19,31}: fix threaded irq backport for <2.6.19
In reply to Arnaud Lacombe--- include/linux/compat-2.6.31.h | 17 +++++++++++++++++ 1 files changed, 17 insertions(+), 0 deletions(-)
include/linux/compat-2.6.31.h +17 -0
@@ -101,6 +101,10 @@ static inline struct rtable *skb_rtable(const struct sk_buff *skb) }/* Backport threaded IRQ support */ +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,19 )) +typedef irqreturn_t (*irq_handler_t )(int ,void *,struct pt_regs *); +#endif +struct compat_threaded_irq {unsigned int irq;irq_handler_t handler;
@@ -117,7 +121,11 @@ void compat_irq_work(void *arg) {struct work_struct *work = arg; struct compat_threaded_irq *comp = container_of(work, struct compat_threaded_irq, work); +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,19 )) + comp->thread_fn (comp->irq, comp->dev_id,NULL ); +#else comp->thread_fn (comp->irq, comp->dev_id); +#endif }#else static inline
@@ -128,13 +136,22 @@ void compat_irq_work(struct work_struct *work) }#endif +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,19 )) +static inline +irqreturn_t compat_irq_dispatcher(int irq, void *dev_id, struct pt_regs *regs) +#else static inline irqreturn_t compat_irq_dispatcher(int irq, void *dev_id) +#endif {struct compat_threaded_irq *comp = dev_id;irqreturn_t res; +#if (LINUX_VERSION_CODE <KERNEL_VERSION (2 ,6 ,19 )) + res = comp->handler (irq, comp->dev_id, regs); +#else res = comp->handler (irq, comp->dev_id); +#endif if (res ==IRQ_WAKE_THREAD ) {queue_work (comp->wq, &comp->work); res =IRQ_HANDLED ;
Signature · 1 line
1.7.2.30.gc37d7.dirty
PATCH 44/44 compat/2.6.35: 2.6.35 does not define `SHRT_MIN'
In reply to Arnaud Lacombe--- include/linux/compat-2.6.35.h | 1 - 1 files changed, 0 insertions(+), 1 deletions(-)
include/linux/compat-2.6.35.h +0 -1
@@ -11,7 +11,6 @@/* added on linux/kernel.h */ #define USHRT_MAX ((u16)(~0U))#define SHRT_MAX ((s16)(USHRT_MAX>>1)) -#define SHRT_MIN ((s16)(-SHRT_MAX - 1))#define SDIO_BUS_ECSI 0x20/* Enable continuous SPI interrupt */ #define SDIO_BUS_SCSI 0x40/* Support continuous SPI interrupt */
Signature · 1 line
1.7.2.30.gc37d7.dirty
Quoted 7 lines · Arnaud Lacombe
Hi Luis,
You will find in answer of this mail the first part of the compat upgrade I am using to support a network-oriented module on kernel from 2.6.15 to 2.6.36.
All this code is available on my repository on github:
Quoted 8 lines
https://github.com/lacombar/linux-compat.git local
This repository also contains the second part I will post later concerning a tentative to reduce namespace pollution when using the compat module. As it is mostly mechanical, I will start another thread with a single proof of concept patch and gather feedbacks :)
Comments welcome !
All in all, very sexy, thanks alot. Can you provide me with a git pull url once you review the comments? If you can first
git fetch; git rebase origin/master
that would be great too.
Thanks a lot!
Luis