Skip to contentSkip to conversation

uml: sparse inlines gcc-4.{2,3} compilation hack

PATCH Sergei Trofimovich 2 participants 2 messages Jul 10, 2008

Subsystems (7)
lore
Share & tools
lore
#0 lore raw

Hi!

Small ARCH=um SUBARCH=i386 .config is in attach. How to fix properly such kind of miscompilation? (gcc-4.3.1) Now I have to use hacks like this:

uml: sparse inlines gcc-4.{2,3} compilation fix.

Patch fixes stuff like this:

CC mm/filemap.o
/mnt/ext/backup3/dev/git/linux-2.6/mm/filemap.c: In function '__generic_file_aio_write_nolock':
/mnt/ext/backup3/dev/git/linux-2.6/mm/filemap.c:1838: sorry, unimplemented: inlining failed in call to 'generic_write_checks': function body not available /mnt/ext/backup3/dev/git/linux-2.6/mm/filemap.c:2390: sorry, unimplemented: called from here
make[2]: *** [mm/filemap.o] Error 1
make[1]: *** [mm] Error 2

 fs/block_dev.c       |    2 +-
 fs/buffer.c          |    2 +-
 mm/filemap.c         |    2 +-
 mm/page_alloc.c      |    2 +-
 net/ipv4/ip_output.c |    2 +-
 5 files changed, 5 insertions(+), 5 deletions(-)
fs/block_dev.c +1 -1
index 10d8a0a..9248ae0 100644
@@ -39,7 +39,7 @@ static inline struct bdev_inode *BDEV_I(struct inode *inode)
        return container_of(inode, struct bdev_inode, vfs_inode);
 }
 
-inline struct block_device *I_BDEV(struct inode *inode)
+struct block_device *I_BDEV(struct inode *inode)
 {
        return &BDEV_I(inode)->bdev;
 }
fs/buffer.c +1 -1
index 0f51c0f..2a9a554 100644
@@ -46,7 +46,7 @@ static int fsync_buffers_list(spinlock_t *lock, struct list_head *list);
 
 #define BH_ENTRY(list) list_entry((list), struct buffer_head, b_assoc_buffers)
 
-inline void
+void
 init_buffer(struct buffer_head *bh, bh_end_io_t *handler, void *private)
 {
        bh->b_end_io = handler;
mm/filemap.c +1 -1
index 1e6a7d3..110b1bc 100644
@@ -1834,7 +1834,7 @@ EXPORT_SYMBOL(iov_iter_single_seg_count);
  * Returns appropriate error code that caller should return or
  * zero in case that write should be allowed.
  */
-inline int generic_write_checks(struct file *file, loff_t *pos, size_t *count, int isblk)
+int generic_write_checks(struct file *file, loff_t *pos, size_t *count, int isblk)
 {
        struct inode *inode = file->f_mapping->host;
        unsigned long limit = current->signal->rlim[RLIMIT_FSIZE].rlim_cur;
mm/page_alloc.c +1 -1
index f32fae3..66b636e 100644
@@ -2611,7 +2611,7 @@ static int zone_batchsize(struct zone *zone)
        return batch;
 }
 
-inline void setup_pageset(struct per_cpu_pageset *p, unsigned long batch)
+void setup_pageset(struct per_cpu_pageset *p, unsigned long batch)
 {
        struct per_cpu_pages *pcp;
 
net/ipv4/ip_output.c +1 -1
index e527628..9a74088 100644
@@ -85,7 +85,7 @@
 int sysctl_ip_default_ttl __read_mostly = IPDEFTTL;
 
 /* Generate a checksum for an outgoing IP datagram. */
-__inline__ void ip_send_check(struct iphdr *iph)
+void ip_send_check(struct iphdr *iph)
 {
        iph->check = 0;
        iph->check = ip_fast_csum((unsigned char *)iph, iph->ihl);
#1 lore raw
Quoted 3 lines · Sergei Trofimovich

Small ARCH=um SUBARCH=i386 .config is in attach. How to fix properly such kind of miscompilation? (gcc-4.3.1) Now I have to use hacks like this:

Use the patch below instead.

Jeff

Signature · 35 lines
Work email - jdike at linux dot intel dot com


commit 4f81c5350b44bcc501ab6f8a089b16d064b4d2f6
Author: Jeff Dike <[email protected]>
Date:   Mon Jul 7 13:36:56 2008 -0400

    [UML] fix gcc ICEs and unresolved externs
    
    There are various constraints on the use of unit-at-a-time:
     - i386 uses no-unit-at-a-time for pre-4.0 (not 4.3)
     - x86_64 uses unit-at-a-time always
    
    Uli reported a crash on x86_64 with gcc 4.1.2 with unit-at-a-time,
    resulting in commit c0a18111e571138747a98af18b3a2124df56a0d1
    
    Ingo reported a gcc internal error with gcc 4.3 with no-unit-at-a-timem,
    resulting in 22eecde2f9034764a3fd095eecfa3adfb8ec9a98
    
    Benny Halevy is seeing extern inlines not resolved with gcc 4.3 with
    no-unit-at-a-time
    
    This patch reintroduces unit-at-a-time for gcc >= 4.0, bringing back the
    possibility of Uli's crash.  If that happens, we'll debug it.
    
    I started seeing both the internal compiler errors and unresolved
    inlines on Fedora 9.  This patch fixes both problems, without so far
    reintroducing the crash reported by Uli.
    
    Signed-off-by: Jeff Dike <[email protected]>
    Cc: Benny Halevy <[email protected]>
    Cc: Adrian Bunk <[email protected]>
    Cc: Ingo Molnar <[email protected]>
    Cc: Ulrich Drepper <[email protected]>
    Signed-off-by: Linus Torvalds <[email protected]>
arch/um/Makefile +0 -1
index dbeab15..ca40397 100644
@@ -77,7 +77,6 @@ include $(srctree)/$(ARCH_DIR)/Makefile-os-$(OS)
 KERNEL_DEFINES = $(strip -Derrno=kernel_errno -Dsigprocmask=kernel_sigprocmask \
 			 -Dmktime=kernel_mktime $(ARCH_KERNEL_DEFINES))
 KBUILD_CFLAGS += $(KERNEL_DEFINES)
-KBUILD_CFLAGS += $(call cc-option,-fno-unit-at-a-time,)
 
 PHONY += linux
 
arch/um/Makefile-i386 +7 -0
index 561e373..302cbe5 100644
@@ -32,4 +32,11 @@ cflags-y += $(call cc-option,-mpreferred-stack-boundary=2)
 # an unresolved reference.
 cflags-y += -ffreestanding
 
+# Disable unit-at-a-time mode on pre-gcc-4.0 compilers, it makes gcc use
+# a lot more stack due to the lack of sharing of stacklots.  Also, gcc
+# 4.3.0 needs -funit-at-a-time for extern inline functions.
+KBUILD_CFLAGS += $(shell if [ $(call cc-version) -lt 0400 ] ; then \
+			echo $(call cc-option,-fno-unit-at-a-time); \
+			else echo $(call cc-option,-funit-at-a-time); fi ;)
+
 KBUILD_CFLAGS += $(cflags-y)
arch/um/Makefile-x86_64 +3 -0
index 8ed362f..a9cd7e7 100644
@@ -21,3 +21,6 @@ HEADER_ARCH := x86
 
 LINK-$(CONFIG_LD_SCRIPT_DYN) += -Wl,-rpath,/lib64
 LINK-y += -m64
+
+# Do unit-at-a-time unconditionally on x86_64, following the host
+KBUILD_CFLAGS += $(call cc-option,-funit-at-a-time)