ia64: fix a few section mismatch warnings
PATCH 1/1 ia64: fix a few section mismatch warnings
Fix the following section mismatch warnings:
WARNING: vmlinux.o(.text+0x41902): Section mismatch: reference to .init.text:__alloc_bootmem (between 'ia64_mca_cpu_init' and 'ia64_do_tlb_purge')
WARNING: vmlinux.o(.text+0x49222): Section mismatch: reference to .init.text:__alloc_bootmem (between 'register_intr' and 'iosapic_register_intr')
WARNING: vmlinux.o(.text+0x62beb2): Section mismatch: reference to .init.text:__alloc_bootmem_node (between 'hubdev_init_node' and 'cnodeid_get_geoid')
For two of the warnings a helper function marked __init_refok was used. For the last warning a missing __init annotation was added.
---
The two remaning warnings in the defconfig build I could not fix:
WARNING: vmlinux.o(.text+0x5b0e2): Section mismatch: reference to .init.text:memmap_init_zone (between 'virtual_memmap_init' and 'memmap_init')
WARNING: vmlinux.o(.text+0x5b182): Section mismatch: reference to .init.text:memmap_init_zone (between 'memmap_init' and 'find_max_min_low_pfn')
I was not sure about the right way to do it since I did not understand the code in mm/init.c deeply enough.
Sam
arch/ia64/kernel/iosapic.c +6 -1
@@ -560,6 +560,11 @@ iosapic_reassign_vector (int irq) } } +static void *__init_refok alloc_rte(unsigned long size) +{ +return alloc_bootmem (size); +} +static struct iosapic_rte_info *iosapic_alloc_rte (void ) {int i;
@@ -567,7 +572,7 @@ static struct iosapic_rte_info *iosapic_alloc_rte (void)int preallocated =0 ;if (!iosapic_kmalloc_ok &&list_empty (&free_rte_list)) { - rte =alloc_bootmem (sizeof (struct iosapic_rte_info ) * + rte =alloc_rte (sizeof (struct iosapic_rte_info ) *NR_PREALLOCATE_RTE_ENTRIES );if (!rte)return NULL ;
arch/ia64/kernel/mca.c +5 -1
@@ -1751,6 +1751,10 @@ format_mca_init_stack(void *mca_data, unsigned long offset, }/* Do per-CPU MCA-related initialization. */ +static void * __init_refok mca_bootmem(unsigned long size) +{ +return alloc_bootmem (size); +}void __cpuinit ia64_mca_cpu_init (void *cpu_data)
@@ -1763,7 +1767,7 @@ ia64_mca_cpu_init(void *cpu_data)int cpu; first_time =0 ; - mca_data =alloc_bootmem (sizeof (struct ia64_mca_cpu ) + mca_data =mca_bootmem (sizeof (struct ia64_mca_cpu ) *NR_CPUS +KERNEL_STACK_SIZE ); mca_data = (void *)(((unsigned long)mca_data + KERNEL_STACK_SIZE - 1) &
arch/ia64/sn/kernel/io_common.c +1 -1
@@ -391,7 +391,7 @@ void sn_bus_free_sysdata(void)* hubdev_init_node() - Creates the HUB data structure and link them to it's * own NODE specific data area. */ -void hubdev_init_node (nodepda_t * npda,cnodeid_t node) +void __init hubdev_init_node (nodepda_t * npda,cnodeid_t node) {struct hubdev_info *hubdev_info;int size;
arch/ia64/kernel/iosapic.c iosapic_reassign_vector (int irq) · line 566
Quoted 4 lines · Sam Ravnborg
arch/ia64/kernel/iosapic.c +4
-0+static void *__init_refok alloc_rte(unsigned long size) +{ +return alloc_bootmem (size); +}
That makes no sense at all. If we ever call that after freeing initmem, we are screwed, period. Sounds like __init fodder.
Quoted 9 lines · Al Viro
arch/ia64/kernel/iosapic.c iosapic_reassign_vector (int irq) · line 566
Quoted 4 lines · Sam Ravnborg
arch/ia64/kernel/iosapic.c +4
-0+static void *__init_refok alloc_rte(unsigned long size) +{ +return alloc_bootmem (size); +}That makes no sense at all. If we ever call that after freeing initmem, we are screwed, period. Sounds like
__initfodder.
The call site has logic to prevent this from being called after init. And the call site cannot be made __init and to limit the scope of the __init_refok a small function is used.
So unless I mis-understood something the above should be OK.
Sam
Quoted 11 lines · Sam Ravnborg
Quoted 9 lines · Al Viro
arch/ia64/kernel/iosapic.c iosapic_reassign_vector (int irq) · line 566
Quoted 4 lines · Sam Ravnborg
arch/ia64/kernel/iosapic.c +4
-0+static void *__init_refok alloc_rte(unsigned long size) +{ +return alloc_bootmem (size); +}That makes no sense at all. If we ever call that after freeing initmem, we are screwed, period. Sounds like
__initfodder.
Quoted 5 lines
The call site has logic to prevent this from being called after init. And the call site cannot be made
__initand to limit the scope of the__init_refoka small function is used.So unless I mis-understood something the above should be OK.
Then the call site must be __init_refok, AFAICS. The point is, the callers of alloc_rte() are where the analysis belongs; use of __init_refok means that you assert that all calls of __init *from* *it* are safe. Otherwise it just becomes "make modpost STFU and let's hope that code is really OK".
Quoted 16 lines · Al Viro
Quoted 15 lines · Sam Ravnborg
Quoted 9 lines · Al Viro
arch/ia64/kernel/iosapic.c iosapic_reassign_vector (int irq) · line 566
Quoted 4 lines · Sam Ravnborg
arch/ia64/kernel/iosapic.c +4
-0+static void *__init_refok alloc_rte(unsigned long size) +{ +return alloc_bootmem (size); +}That makes no sense at all. If we ever call that after freeing initmem, we are screwed, period. Sounds like
__initfodder.The call site has logic to prevent this from being called after init. And the call site cannot be made
__initand to limit the scope of the__init_refoka small function is used.
Quoted 7 lines
Quoted 1 line
So unless I mis-understood something the above should be OK.
Then the call site must be
__init_refok, AFAICS. The point is, the callers ofalloc_rte()are where the analysis belongs; use of__init_refokmeans that you assert that all calls of__init*from* *it* are safe. Otherwise it just becomes "make modpost STFU and let's hope that code is really OK".
The above function is suposed to be used solely from iosapic_alloc_rte() and not for general use.
And instead of declaring all of iosapic_alloc_rte() as __init_refok then the single line doing the call to an __init function was factored out and marked __init.alloc_rte() can be used from everywhere but the naming suggest that this is for rte only.
To make it more clear what the function was used for I could have made it like this:
/*
* Call site shall make sure this isused only during early init.
* Use __init_refok to avoid warnings from modpost.
*/
static void * __init_refok alloc_rte(unsigned long size)
{
return alloc_bootmen(sizeof(struct iosapic_rte_info) * size);
}Would that make the intention more clear?
As for marking all of iosapic_alloc_rte() __init_refok this was not done because that would then not detect any new __init usage within the function. That could be done too obviously - it is anyway only 30 loc.
Sam
Fix the following section mismatch warnings:
WARNING: vmlinux.o(.text+0x41902): Section mismatch: reference to .init.text:__alloc_bootmem (between 'ia64_mca_cpu_init' and 'ia64_do_tlb_purge')
WARNING: vmlinux.o(.text+0x49222): Section mismatch: reference to .init.text:__alloc_bootmem (between 'register_intr' and 'iosapic_register_intr')
WARNING: vmlinux.o(.text+0x62beb2): Section mismatch: reference to .init.text:__alloc_bootmem_node (between 'hubdev_init_node' and 'cnodeid_get_geoid')
For ithe first warning a helper function marked __init_refok was used. For the last warning a missing __init annotation was added.
---
Updated - inspired by comments from Al Viro.
arch/ia64/kernel/iosapic.c +1 -1
@@ -560,7 +560,7 @@ iosapic_reassign_vector (int irq) } } -static struct iosapic_rte_info *iosapic_alloc_rte (void ) +static struct iosapic_rte_info * __init_refok iosapic_alloc_rte (void) {int i;struct iosapic_rte_info *rte;
arch/ia64/kernel/mca.c +5 -2
@@ -1751,6 +1751,10 @@ format_mca_init_stack(void *mca_data, unsigned long offset, }/* Do per-CPU MCA-related initialization. */ +static void * __init_refok mca_bootmem(unsigned long size) +{ +return alloc_bootmem (size *sizeof (struct ia64_mca_cpu )); +}void __cpuinit ia64_mca_cpu_init (void *cpu_data)
@@ -1763,8 +1767,7 @@ ia64_mca_cpu_init(void *cpu_data)int cpu; first_time =0 ; - mca_data =alloc_bootmem (sizeof (struct ia64_mca_cpu ) - *NR_CPUS +KERNEL_STACK_SIZE ); + mca_data =mca_bootmem (NR_CPUS +KERNEL_STACK_SIZE ); mca_data = (void *)(((unsigned long )mca_data +KERNEL_STACK_SIZE -1 ) & (-KERNEL_STACK_SIZE ));
arch/ia64/sn/kernel/io_common.c +1 -1
@@ -391,7 +391,7 @@ void sn_bus_free_sysdata(void)* hubdev_init_node() - Creates the HUB data structure and link them to it's * own NODE specific data area. */ -void hubdev_init_node (nodepda_t * npda,cnodeid_t node) +void __init hubdev_init_node (nodepda_t * npda,cnodeid_t node) {struct hubdev_info *hubdev_info;int size;
- mca_data = alloc_bootmem(sizeof(struct ia64_mca_cpu)
- * NR_CPUS + KERNEL_STACK_SIZE);
+ mca_data = mca_bootmem(NR_CPUS + KERNEL_STACK_SIZE);
Oops. You moved the multiply by sizeof(struct ia64_mca_cpu) up into the mca_bootmem() function to make it very specific to this use. But mutiply has higher precedence than addition.
-Tony
Quoted 7 lines · Tony
arch/ia64/kernel/mca.c +1
-2- mca_data =alloc_bootmem (sizeof (struct ia64_mca_cpu ) - *NR_CPUS +KERNEL_STACK_SIZE ); + mca_data =mca_bootmem (NR_CPUS +KERNEL_STACK_SIZE );Oops. You moved the multiply by
sizeof(struct ia64_mca_cpu)up into themca_bootmem()function to make it very specific to this use. But mutiply has higher precedence than addition.
Oh crap - good catch.
Shall I resubmit a corrected patch?
Sam
Quoted 6 lines
Quoted 3 lines
Oops. You moved the multiply by
sizeof(struct ia64_mca_cpu)up into themca_bootmem()function to make it very specific to this use. But mutiply has higher precedence than addition.Oh crap - good catch.
Shall I resubmit a corrected patch?
Are there any other ways that we might tag the callsite to let modpost know that this instance is safe? Adding a call to a wrapper function in __init_refok space feels kludgy, and whatever comments you stick on that function, it is sitting there waiting for someone who shouldn't to make a call.
If not ... then just make the mca_bootmem() function take no args. It can calculate the amount of memory, it can do the next bit too and return a result KERNEL_STACK_SIZE aligned too.
-Tony
PATCH v3 1/1 ia64: fix a few section mismatch warnings
Fix the following section mismatch warnings:
WARNING: vmlinux.o(.text+0x41902): Section mismatch: reference to .init.text:__alloc_bootmem (between 'ia64_mca_cpu_init' and 'ia64_do_tlb_purge')
WARNING: vmlinux.o(.text+0x49222): Section mismatch: reference to .init.text:__alloc_bootmem (between 'register_intr' and 'iosapic_register_intr')
WARNING: vmlinux.o(.text+0x62beb2): Section mismatch: reference to .init.text:__alloc_bootmem_node (between 'hubdev_init_node' and 'cnodeid_get_geoid')
---
This should be in line with your latest comments. Notice I introduce use of ALIGN to make it more readable. I actually made a smal .c program just to verify that the below expression and ALIGN agreed on how to calculate the alignment.
Sam
arch/ia64/kernel/iosapic.c +1 -1
@@ -560,7 +560,7 @@ iosapic_reassign_vector (int irq) } } -static struct iosapic_rte_info *iosapic_alloc_rte (void ) +static struct iosapic_rte_info * __init_refok iosapic_alloc_rte (void) {int i;struct iosapic_rte_info *rte;
arch/ia64/kernel/mca.c +11 -6
@@ -1750,8 +1750,17 @@ format_mca_init_stack(void *mca_data, unsigned long offset,strncpy (p->comm, type,sizeof (p->comm)-1 ); } -/* Do per-CPU MCA-related initialization. */ +/* Caller prevents this from being called after init */ +static void * __init_refok mca_bootmem(void) +{ +void *p; + p =alloc_bootmem (sizeof (struct ia64_mca_cpu ) *NR_CPUS + +KERNEL_STACK_SIZE ); +return (void *)ALIGN ((unsigned long )p,KERNEL_STACK_SIZE ); +} + +/* Do per-CPU MCA-related initialization. */ void __cpuinit ia64_mca_cpu_init (void *cpu_data) {
@@ -1763,11 +1772,7 @@ ia64_mca_cpu_init(void *cpu_data)int cpu; first_time =0 ; - mca_data =alloc_bootmem (sizeof (struct ia64_mca_cpu ) - *NR_CPUS +KERNEL_STACK_SIZE ); - mca_data = (void *)(((unsigned long )mca_data + -KERNEL_STACK_SIZE -1 ) & - (-KERNEL_STACK_SIZE )); + mca_data =mca_bootmem ();for (cpu =0 ; cpu <NR_CPUS ; cpu++) { format_mca_init_stack(mca_data, offsetof(struct ia64_mca_cpu, mca_stack),
arch/ia64/sn/kernel/io_common.c +1 -1
@@ -391,7 +391,7 @@ void sn_bus_free_sysdata(void)* hubdev_init_node() - Creates the HUB data structure and link them to it's * own NODE specific data area. */ -void hubdev_init_node (nodepda_t * npda,cnodeid_t node) +void __init hubdev_init_node (nodepda_t * npda,cnodeid_t node) {struct hubdev_info *hubdev_info;int size;