x86/entry: Create a home for the x86 entry code in arch/x86/entry/

So the x86 syscall/irq/etc. entry code is scattered in over 40 files all over the x86 architecture, making it hard to get a good overview of the code and its current status.

Move all the files to arch/x86/entry/.

This first step is as-is, no file names were changed - but the next step will be to organize things in a bit more maintainable fashion.

If there are no fundamental objections then I'll commit this to tip:x86/asm. (I've drained most of the pending patches to these files - but porting them should be easy in any case.)

Thanks,

Ingo

=========>

Ingo Molnar (7):
x86/asm/entry: Move entry_64.S and entry_32.S to arch/x86/entry/ x86/asm/entry: Move the compat syscall entry code to arch/x86/entry/ x86/asm/entry, x86/vdso: Move the vDSO code to arch/x86/entry/vdso/ x86/asm/entry: Move the 'thunk' functions to arch/x86/entry/ x86/asm/entry: Move arch/x86/include/asm/calling.h to arch/x86/entry/ x86/asm/entry: Move the arch/x86/syscalls/ definitions to arch/x86/entry/syscalls/ x86/asm/entry: Move the vsyscall code to arch/x86/entry/vsyscall/

 MAINTAINERS                                           |  2 +-
 arch/x86/Kbuild                                       |  5 ++++-
 arch/x86/Makefile                                     |  4 ++--
 arch/x86/entry/Makefile                               | 10 ++++++++++
 arch/x86/{include/asm => entry}/calling.h             |  0
 arch/x86/{kernel => entry}/entry_32.S                 |  0
 arch/x86/{kernel => entry}/entry_64.S                 |  2 +-
 arch/x86/{ia32 => entry}/ia32entry.S                  |  2 +-
 arch/x86/{kernel => entry}/syscall_32.c               |  0
 arch/x86/{kernel => entry}/syscall_64.c               |  0
 arch/x86/{ => entry}/syscalls/Makefile                |  4 ++--
 arch/x86/{ => entry}/syscalls/syscall_32.tbl          |  0
 arch/x86/{ => entry}/syscalls/syscall_64.tbl          |  0
 arch/x86/{ => entry}/syscalls/syscallhdr.sh           |  0
 arch/x86/{ => entry}/syscalls/syscalltbl.sh           |  0
 arch/x86/{lib => entry}/thunk_32.S                    |  0
 arch/x86/{lib => entry}/thunk_64.S                    |  2 +-
 arch/x86/{ => entry}/vdso/.gitignore                  |  0
 arch/x86/{ => entry}/vdso/Makefile                    |  0
 arch/x86/{ => entry}/vdso/checkundef.sh               |  0
 arch/x86/{ => entry}/vdso/vclock_gettime.c            |  0
 arch/x86/{ => entry}/vdso/vdso-layout.lds.S           |  0
 arch/x86/{ => entry}/vdso/vdso-note.S                 |  0
 arch/x86/{ => entry}/vdso/vdso.lds.S                  |  0
 arch/x86/{ => entry}/vdso/vdso2c.c                    |  0
 arch/x86/{ => entry}/vdso/vdso2c.h                    |  0
 arch/x86/{ => entry}/vdso/vdso32-setup.c              |  0
 arch/x86/{ => entry}/vdso/vdso32/.gitignore           |  0
 arch/x86/{ => entry}/vdso/vdso32/int80.S              |  0
 arch/x86/{ => entry}/vdso/vdso32/note.S               |  0
 arch/x86/{ => entry}/vdso/vdso32/sigreturn.S          |  0
 arch/x86/{ => entry}/vdso/vdso32/syscall.S            |  0
 arch/x86/{ => entry}/vdso/vdso32/sysenter.S           |  0
 arch/x86/{ => entry}/vdso/vdso32/vclock_gettime.c     |  0
 arch/x86/{ => entry}/vdso/vdso32/vdso-fakesections.c  |  0
 arch/x86/{ => entry}/vdso/vdso32/vdso32.lds.S         |  0
 arch/x86/{ => entry}/vdso/vdsox32.lds.S               |  0
 arch/x86/{ => entry}/vdso/vgetcpu.c                   |  0
 arch/x86/{ => entry}/vdso/vma.c                       |  0
 arch/x86/entry/vsyscall/Makefile                      |  7 +++++++
 arch/x86/{kernel => entry/vsyscall}/vsyscall_64.c     |  0
 arch/x86/{kernel => entry/vsyscall}/vsyscall_emu_64.S |  0
 arch/x86/{kernel => entry/vsyscall}/vsyscall_gtod.c   |  0
 arch/x86/{kernel => entry/vsyscall}/vsyscall_trace.h  |  2 +-
 arch/x86/ia32/Makefile                                |  2 +-
 arch/x86/kernel/Makefile                              |  5 +----
 arch/x86/lib/Makefile                                 |  1 -
 scripts/checksyscalls.sh                              |  2 +-
 48 files changed, 33 insertions(+), 17 deletions(-)
 create mode 100644 arch/x86/entry/Makefile
 rename arch/x86/{include/asm => entry}/calling.h (100%)
 rename arch/x86/{kernel => entry}/entry_32.S (100%)
 rename arch/x86/{kernel => entry}/entry_64.S (99%)
 rename arch/x86/{ia32 => entry}/ia32entry.S (99%)
 rename arch/x86/{kernel => entry}/syscall_32.c (100%)
 rename arch/x86/{kernel => entry}/syscall_64.c (100%)
 rename arch/x86/{ => entry}/syscalls/Makefile (95%)
 rename arch/x86/{ => entry}/syscalls/syscall_32.tbl (100%)
 rename arch/x86/{ => entry}/syscalls/syscall_64.tbl (100%)
 rename arch/x86/{ => entry}/syscalls/syscallhdr.sh (100%)
 rename arch/x86/{ => entry}/syscalls/syscalltbl.sh (100%)
 rename arch/x86/{lib => entry}/thunk_32.S (100%)
 rename arch/x86/{lib => entry}/thunk_64.S (98%)
 rename arch/x86/{ => entry}/vdso/.gitignore (100%)
 rename arch/x86/{ => entry}/vdso/Makefile (100%)
 rename arch/x86/{ => entry}/vdso/checkundef.sh (100%)
 rename arch/x86/{ => entry}/vdso/vclock_gettime.c (100%)
 rename arch/x86/{ => entry}/vdso/vdso-layout.lds.S (100%)
 rename arch/x86/{ => entry}/vdso/vdso-note.S (100%)
 rename arch/x86/{ => entry}/vdso/vdso.lds.S (100%)
 rename arch/x86/{ => entry}/vdso/vdso2c.c (100%)
 rename arch/x86/{ => entry}/vdso/vdso2c.h (100%)
 rename arch/x86/{ => entry}/vdso/vdso32-setup.c (100%)
 rename arch/x86/{ => entry}/vdso/vdso32/.gitignore (100%)
 rename arch/x86/{ => entry}/vdso/vdso32/int80.S (100%)
 rename arch/x86/{ => entry}/vdso/vdso32/note.S (100%)
 rename arch/x86/{ => entry}/vdso/vdso32/sigreturn.S (100%)
 rename arch/x86/{ => entry}/vdso/vdso32/syscall.S (100%)
 rename arch/x86/{ => entry}/vdso/vdso32/sysenter.S (100%)
 rename arch/x86/{ => entry}/vdso/vdso32/vclock_gettime.c (100%)
 rename arch/x86/{ => entry}/vdso/vdso32/vdso-fakesections.c (100%)
 rename arch/x86/{ => entry}/vdso/vdso32/vdso32.lds.S (100%)
 rename arch/x86/{ => entry}/vdso/vdsox32.lds.S (100%)
 rename arch/x86/{ => entry}/vdso/vgetcpu.c (100%)
 rename arch/x86/{ => entry}/vdso/vma.c (100%)
 create mode 100644 arch/x86/entry/vsyscall/Makefile
 rename arch/x86/{kernel => entry/vsyscall}/vsyscall_64.c (100%)
 rename arch/x86/{kernel => entry/vsyscall}/vsyscall_emu_64.S (100%)
 rename arch/x86/{kernel => entry/vsyscall}/vsyscall_gtod.c (100%)
 rename arch/x86/{kernel => entry/vsyscall}/vsyscall_trace.h (89%)

2.1.4

Create a new directory hierarchy for the low level x86 entry code:

    arch/x86/entry/*

This will host all the low level glue that is currently scattered all across arch/x86/.

Start with entry_64.S and entry_32.S.

---
 arch/x86/Kbuild                       | 3 +++
 arch/x86/entry/Makefile               | 4 ++++
 arch/x86/{kernel => entry}/entry_32.S | 0
 arch/x86/{kernel => entry}/entry_64.S | 0
 arch/x86/kernel/Makefile              | 2 +-
 5 files changed, 8 insertions(+), 1 deletion(-)

index 3942f74c92d7..b9b816277e72 100644
@@ -1,3 +1,6 @@
+
+obj-y += entry/
+
 obj-$(CONFIG_KVM) += kvm/
 
 # Xen paravirtualization support

new file mode 100644
index 000000000000..fa7e0cf6d3c4
@@ -0,0 +1,4 @@
+#
+# Makefile for the x86 low level entry code
+#
+obj-y			:= entry_$(BITS).o

similarity index 100%
rename from arch/x86/kernel/entry_32.S
rename to arch/x86/entry/entry_32.S

similarity index 100%
rename from arch/x86/kernel/entry_64.S
rename to arch/x86/entry/entry_64.S

index 9bcd0b56ca17..9d3ee054453d 100644
@@ -22,7 +22,7 @@ KASAN_SANITIZE_dumpstack_$(BITS).o := n
 
 CFLAGS_irq.o := -I$(src)/../include/asm/trace
 
-obj-y			:= process_$(BITS).o signal.o entry_$(BITS).o
+obj-y			:= process_$(BITS).o signal.o
 obj-y			+= traps.o irq.o irq_$(BITS).o dumpstack_$(BITS).o
 obj-y			+= time.o ioport.o ldt.o dumpstack.o nmi.o
 obj-y			+= setup.o x86_init.o i8259.o irqinit.o jump_label.o

2.1.4

Move the ia32entry.S file over into arch/x86/entry/.

---
 arch/x86/entry/Makefile              | 5 ++++-
 arch/x86/{ia32 => entry}/ia32entry.S | 0
 arch/x86/ia32/Makefile               | 2 +-
 3 files changed, 5 insertions(+), 2 deletions(-)

index fa7e0cf6d3c4..4a626594fa79 100644
@@ -1,4 +1,7 @@
 #
 # Makefile for the x86 low level entry code
 #
-obj-y			:= entry_$(BITS).o
+obj-y				:= entry_$(BITS).o
+
+obj-$(CONFIG_IA32_EMULATION)	+= ia32entry.o
+

similarity index 100%
rename from arch/x86/ia32/ia32entry.S
rename to arch/x86/entry/ia32entry.S

index bb635c641869..cd4339bae066 100644
@@ -2,7 +2,7 @@
 # Makefile for the ia32 kernel emulation subsystem.
 #
 
-obj-$(CONFIG_IA32_EMULATION) := ia32entry.o sys_ia32.o ia32_signal.o
+obj-$(CONFIG_IA32_EMULATION) := sys_ia32.o ia32_signal.o
 
 obj-$(CONFIG_IA32_AOUT) += ia32_aout.o
 

2.1.4

I think we should also rename it to compat_entry.S or something. ia32entry sounds suspiciously like 32-bit code.

--Andy

Absolutely, I alluded to that in the 0/7 description, but wanted to get the 'plain' movement done as a first step.

Thanks,

Ingo

---
 MAINTAINERS                                          | 2 +-
 arch/x86/Kbuild                                      | 2 +-
 arch/x86/Makefile                                    | 2 +-
 arch/x86/entry/Makefile                              | 2 ++
 arch/x86/{ => entry}/vdso/.gitignore                 | 0
 arch/x86/{ => entry}/vdso/Makefile                   | 0
 arch/x86/{ => entry}/vdso/checkundef.sh              | 0
 arch/x86/{ => entry}/vdso/vclock_gettime.c           | 0
 arch/x86/{ => entry}/vdso/vdso-layout.lds.S          | 0
 arch/x86/{ => entry}/vdso/vdso-note.S                | 0
 arch/x86/{ => entry}/vdso/vdso.lds.S                 | 0
 arch/x86/{ => entry}/vdso/vdso2c.c                   | 0
 arch/x86/{ => entry}/vdso/vdso2c.h                   | 0
 arch/x86/{ => entry}/vdso/vdso32-setup.c             | 0
 arch/x86/{ => entry}/vdso/vdso32/.gitignore          | 0
 arch/x86/{ => entry}/vdso/vdso32/int80.S             | 0
 arch/x86/{ => entry}/vdso/vdso32/note.S              | 0
 arch/x86/{ => entry}/vdso/vdso32/sigreturn.S         | 0
 arch/x86/{ => entry}/vdso/vdso32/syscall.S           | 0
 arch/x86/{ => entry}/vdso/vdso32/sysenter.S          | 0
 arch/x86/{ => entry}/vdso/vdso32/vclock_gettime.c    | 0
 arch/x86/{ => entry}/vdso/vdso32/vdso-fakesections.c | 0
 arch/x86/{ => entry}/vdso/vdso32/vdso32.lds.S        | 0
 arch/x86/{ => entry}/vdso/vdsox32.lds.S              | 0
 arch/x86/{ => entry}/vdso/vgetcpu.c                  | 0
 arch/x86/{ => entry}/vdso/vma.c                      | 0
 26 files changed, 5 insertions(+), 3 deletions(-)

index f8e0afb708b4..68c0cc365432 100644
@@ -10892,7 +10892,7 @@ M:	Andy Lutomirski <[email protected]>
 L:	[email protected]
 T:	git git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git x86/vdso
 S:	Maintained
-F:	arch/x86/vdso/
+F:	arch/x86/entry/vdso/
 
 XC2028/3028 TUNER DRIVER
 M:	Mauro Carvalho Chehab <[email protected]>

index b9b816277e72..1538562cc720 100644
@@ -14,7 +14,7 @@ obj-y += kernel/
 obj-y += mm/
 
 obj-y += crypto/
-obj-y += vdso/
+
 obj-$(CONFIG_IA32_EMULATION) += ia32/
 
 obj-y += platform/

index 43e8328a23e4..90b1c3bfec46 100644
@@ -244,7 +244,7 @@ PHONY += install
 
 PHONY += vdso_install
 vdso_install:
-	$(Q)$(MAKE) $(build)=arch/x86/vdso $@
+	$(Q)$(MAKE) $(build)=arch/x86/entry/vdso $@
 
 archclean:
 	$(Q)rm -rf $(objtree)/arch/i386

index 4a626594fa79..c97532492ef5 100644
@@ -3,5 +3,7 @@
 #
 obj-y				:= entry_$(BITS).o
 
+obj-y				+= vdso/
+
 obj-$(CONFIG_IA32_EMULATION)	+= ia32entry.o
 

similarity index 100%
rename from arch/x86/vdso/.gitignore
rename to arch/x86/entry/vdso/.gitignore

similarity index 100%
rename from arch/x86/vdso/Makefile
rename to arch/x86/entry/vdso/Makefile

similarity index 100%
rename from arch/x86/vdso/checkundef.sh
rename to arch/x86/entry/vdso/checkundef.sh

similarity index 100%
rename from arch/x86/vdso/vclock_gettime.c
rename to arch/x86/entry/vdso/vclock_gettime.c

similarity index 100%
rename from arch/x86/vdso/vdso-layout.lds.S
rename to arch/x86/entry/vdso/vdso-layout.lds.S

similarity index 100%
rename from arch/x86/vdso/vdso-note.S
rename to arch/x86/entry/vdso/vdso-note.S

similarity index 100%
rename from arch/x86/vdso/vdso.lds.S
rename to arch/x86/entry/vdso/vdso.lds.S

similarity index 100%
rename from arch/x86/vdso/vdso2c.c
rename to arch/x86/entry/vdso/vdso2c.c

similarity index 100%
rename from arch/x86/vdso/vdso2c.h
rename to arch/x86/entry/vdso/vdso2c.h

similarity index 100%
rename from arch/x86/vdso/vdso32-setup.c
rename to arch/x86/entry/vdso/vdso32-setup.c

similarity index 100%
rename from arch/x86/vdso/vdso32/.gitignore
rename to arch/x86/entry/vdso/vdso32/.gitignore

similarity index 100%
rename from arch/x86/vdso/vdso32/int80.S
rename to arch/x86/entry/vdso/vdso32/int80.S

similarity index 100%
rename from arch/x86/vdso/vdso32/note.S
rename to arch/x86/entry/vdso/vdso32/note.S

similarity index 100%
rename from arch/x86/vdso/vdso32/sigreturn.S
rename to arch/x86/entry/vdso/vdso32/sigreturn.S

similarity index 100%
rename from arch/x86/vdso/vdso32/syscall.S
rename to arch/x86/entry/vdso/vdso32/syscall.S

similarity index 100%
rename from arch/x86/vdso/vdso32/sysenter.S
rename to arch/x86/entry/vdso/vdso32/sysenter.S

similarity index 100%
rename from arch/x86/vdso/vdso32/vclock_gettime.c
rename to arch/x86/entry/vdso/vdso32/vclock_gettime.c

similarity index 100%
rename from arch/x86/vdso/vdso32/vdso-fakesections.c
rename to arch/x86/entry/vdso/vdso32/vdso-fakesections.c

similarity index 100%
rename from arch/x86/vdso/vdso32/vdso32.lds.S
rename to arch/x86/entry/vdso/vdso32/vdso32.lds.S

similarity index 100%
rename from arch/x86/vdso/vdsox32.lds.S
rename to arch/x86/entry/vdso/vdsox32.lds.S

similarity index 100%
rename from arch/x86/vdso/vgetcpu.c
rename to arch/x86/entry/vdso/vgetcpu.c

similarity index 100%
rename from arch/x86/vdso/vma.c
rename to arch/x86/entry/vdso/vma.c

2.1.4

2015-06-03 9:59 GMT-07:00 Ingo Molnar <[email protected]>:

I'm with hpa here. The vdso isn't really entry code.

--Andy

These are all calling x86 entry code functions, so move them close to other entry code.

Change lib-y to obj-y: there's no real difference between the two as we don't really drop any of them during the linking stage, and obj-y is the more common approach for core kernel object code.

---
 arch/x86/entry/Makefile            | 3 +--
 arch/x86/{lib => entry}/thunk_32.S | 0
 arch/x86/{lib => entry}/thunk_64.S | 0
 arch/x86/lib/Makefile              | 1 -
 4 files changed, 1 insertion(+), 3 deletions(-)

index c97532492ef5..9df72c8a0ac2 100644
@@ -1,8 +1,7 @@
 #
 # Makefile for the x86 low level entry code
 #
-obj-y				:= entry_$(BITS).o
-
+obj-y				:= entry_$(BITS).o thunk_$(BITS).o
 obj-y				+= vdso/
 
 obj-$(CONFIG_IA32_EMULATION)	+= ia32entry.o

similarity index 100%
rename from arch/x86/lib/thunk_32.S
rename to arch/x86/entry/thunk_32.S

similarity index 100%
rename from arch/x86/lib/thunk_64.S
rename to arch/x86/entry/thunk_64.S

index 982989d282ff..f2587888d987 100644
@@ -17,7 +17,6 @@ clean-files := inat-tables.c
 obj-$(CONFIG_SMP) += msr-smp.o cache-smp.o
 
 lib-y := delay.o misc.o cmdline.o
-lib-y += thunk_$(BITS).o
 lib-y += usercopy_$(BITS).o usercopy.o getuser.o putuser.o
 lib-y += memcpy_$(BITS).o
 lib-$(CONFIG_RWSEM_XCHGADD_ALGORITHM) += rwsem.o

2.1.4

asm/calling.h is private to the entry code, make this more apparent by moving it to the new arch/x86/entry/ directory.

---
 arch/x86/{include/asm => entry}/calling.h | 0
 arch/x86/entry/entry_64.S                 | 2 +-
 arch/x86/entry/ia32entry.S                | 2 +-
 arch/x86/entry/thunk_64.S                 | 2 +-
 4 files changed, 3 insertions(+), 3 deletions(-)

similarity index 100%
rename from arch/x86/include/asm/calling.h
rename to arch/x86/entry/calling.h

index 4ad79e946f5a..f7380ea75777 100644
@@ -28,7 +28,7 @@
 #include <asm/segment.h>
 #include <asm/cache.h>
 #include <asm/errno.h>
-#include <asm/calling.h>
+#include "calling.h"
 #include <asm/asm-offsets.h>
 #include <asm/msr.h>
 #include <asm/unistd.h>

index 2be23c734db5..f16767417385 100644
@@ -4,7 +4,7 @@
  * Copyright 2000-2002 Andi Kleen, SuSE Labs.
  */		 
 
-#include <asm/calling.h>
+#include "calling.h"
 #include <asm/asm-offsets.h>
 #include <asm/current.h>
 #include <asm/errno.h>

index 10f555e435e1..3e95681b4e2d 100644
@@ -6,7 +6,7 @@
  * Subject to the GNU public license, v.2. No warranty of any kind.
  */
 #include <linux/linkage.h>
-#include <asm/calling.h>
+#include "calling.h"
 #include <asm/asm.h>
 
 	/* rdi:	arg1 ... normal C conventions. rax is saved/restored. */

2.1.4

The build time generated syscall definitions are entry code related, move them into the arch/x86/entry/ directory.

---
 arch/x86/Makefile                            | 2 +-
 arch/x86/{ => entry}/syscalls/Makefile       | 4 ++--
 arch/x86/{ => entry}/syscalls/syscall_32.tbl | 0
 arch/x86/{ => entry}/syscalls/syscall_64.tbl | 0
 arch/x86/{ => entry}/syscalls/syscallhdr.sh  | 0
 arch/x86/{ => entry}/syscalls/syscalltbl.sh  | 0
 scripts/checksyscalls.sh                     | 2 +-
 7 files changed, 4 insertions(+), 4 deletions(-)

index 90b1c3bfec46..118e6debc483 100644
@@ -181,7 +181,7 @@ archscripts: scripts_basic
 # Syscall table generation
 
 archheaders:
-	$(Q)$(MAKE) $(build)=arch/x86/syscalls all
+	$(Q)$(MAKE) $(build)=arch/x86/entry/syscalls all
 
 archprepare:
 ifeq ($(CONFIG_KEXEC_FILE),y)

similarity index 95%
rename from arch/x86/syscalls/Makefile
rename to arch/x86/entry/syscalls/Makefile
index a55abb9f6c5e..57aa59fd140c 100644
@@ -1,5 +1,5 @@
-out := $(obj)/../include/generated/asm
-uapi := $(obj)/../include/generated/uapi/asm
+out := $(obj)/../../include/generated/asm
+uapi := $(obj)/../../include/generated/uapi/asm
 
 # Create output directory if not already present
 _dummy := $(shell [ -d '$(out)' ] || mkdir -p '$(out)') \

similarity index 100%
rename from arch/x86/syscalls/syscall_32.tbl
rename to arch/x86/entry/syscalls/syscall_32.tbl

similarity index 100%
rename from arch/x86/syscalls/syscall_64.tbl
rename to arch/x86/entry/syscalls/syscall_64.tbl

similarity index 100%
rename from arch/x86/syscalls/syscallhdr.sh
rename to arch/x86/entry/syscalls/syscallhdr.sh

similarity index 100%
rename from arch/x86/syscalls/syscalltbl.sh
rename to arch/x86/entry/syscalls/syscalltbl.sh

index 5b3add31f9f1..2c9082ba6137 100755
@@ -212,5 +212,5 @@ EOF
     )
 }
 
-(ignore_list && syscall_list $(dirname $0)/../arch/x86/syscalls/syscall_32.tbl) | \
+(ignore_list && syscall_list $(dirname $0)/../arch/x86/entry/syscalls/syscall_32.tbl) | \
 $* -E -x c - > /dev/null

2.1.4

The vsyscall code is entry code too, so move it to arch/x86/entry/vsyscall/.

---
 arch/x86/entry/Makefile                               | 6 ++++--
 arch/x86/{kernel => entry}/syscall_32.c               | 0
 arch/x86/{kernel => entry}/syscall_64.c               | 0
 arch/x86/entry/vsyscall/Makefile                      | 7 +++++++
 arch/x86/{kernel => entry/vsyscall}/vsyscall_64.c     | 0
 arch/x86/{kernel => entry/vsyscall}/vsyscall_emu_64.S | 0
 arch/x86/{kernel => entry/vsyscall}/vsyscall_gtod.c   | 0
 arch/x86/{kernel => entry/vsyscall}/vsyscall_trace.h  | 2 +-
 arch/x86/kernel/Makefile                              | 3 ---
 9 files changed, 12 insertions(+), 6 deletions(-)

index 9df72c8a0ac2..b93cce1c6bb2 100644
@@ -1,8 +1,10 @@
 #
 # Makefile for the x86 low level entry code
 #
-obj-y				:= entry_$(BITS).o thunk_$(BITS).o
+obj-y				:= entry_$(BITS).o thunk_$(BITS).o syscall_$(BITS).o
+
 obj-y				+= vdso/
+obj-y				+= vsyscall/
 
-obj-$(CONFIG_IA32_EMULATION)	+= ia32entry.o
+obj-$(CONFIG_IA32_EMULATION)	+= ia32entry.o syscall_32.o
 

similarity index 100%
rename from arch/x86/kernel/syscall_32.c
rename to arch/x86/entry/syscall_32.c

similarity index 100%
rename from arch/x86/kernel/syscall_64.c
rename to arch/x86/entry/syscall_64.c

new file mode 100644
index 000000000000..a9f4856f622a
@@ -0,0 +1,7 @@
+#
+# Makefile for the x86 low level vsyscall code
+#
+obj-y					:= vsyscall_gtod.o
+
+obj-$(CONFIG_X86_VSYSCALL_EMULATION)	+= vsyscall_64.o vsyscall_emu_64.o
+

similarity index 100%
rename from arch/x86/kernel/vsyscall_64.c
rename to arch/x86/entry/vsyscall/vsyscall_64.c

similarity index 100%
rename from arch/x86/kernel/vsyscall_emu_64.S
rename to arch/x86/entry/vsyscall/vsyscall_emu_64.S

similarity index 100%
rename from arch/x86/kernel/vsyscall_gtod.c
rename to arch/x86/entry/vsyscall/vsyscall_gtod.c

similarity index 89%
rename from arch/x86/kernel/vsyscall_trace.h
rename to arch/x86/entry/vsyscall/vsyscall_trace.h
index a8b2edec54fe..9dd7359a38a8 100644
@@ -24,6 +24,6 @@ TRACE_EVENT(emulate_vsyscall,
 #endif
 
 #undef TRACE_INCLUDE_PATH
-#define TRACE_INCLUDE_PATH ../../arch/x86/kernel
+#define TRACE_INCLUDE_PATH ../../arch/x86/entry/vsyscall/
 #define TRACE_INCLUDE_FILE vsyscall_trace
 #include <trace/define_trace.h>

index 9d3ee054453d..01663ee5f1b7 100644
@@ -31,9 +31,6 @@ obj-y			+= probe_roms.o
 obj-$(CONFIG_X86_32)	+= i386_ksyms_32.o
 obj-$(CONFIG_X86_64)	+= sys_x86_64.o x8664_ksyms_64.o
 obj-$(CONFIG_X86_64)	+= mcount_64.o
-obj-y			+= syscall_$(BITS).o vsyscall_gtod.o
-obj-$(CONFIG_IA32_EMULATION)	+= syscall_32.o
-obj-$(CONFIG_X86_VSYSCALL_EMULATION)	+= vsyscall_64.o vsyscall_emu_64.o
 obj-$(CONFIG_X86_ESPFIX64)	+= espfix_64.o
 obj-$(CONFIG_SYSFS)	+= ksysfs.o
 obj-y			+= bootflag.o e820.o

2.1.4

I would prefer if it was made a separate topic.

I like the patch set (and you can add my Acked-by:) *except* 7/7, and the reason for that is that it really isn't entry code, it is user space code.

I also dislike deep directory hierarchies in general, unless it really be avoided, like boot/ or realmode/.

-hpa

Well, I think arch/x86/entry/ should be a broader category for all things entry
code: and the vsyscall code is closely related to the syscall entry/exit code so it's in a better place there than just being in the generic arch/x86/kernel/ directory.

I kept it separate in arch/x86/entry/vsyscall/ so it doesn't mix with other entry code.

Thanks,

Ingo

...and my reading comprehension is way off this morning. You already called it arch/x86/entry, so there was no reason for me to suggest that :)

Anyway, arch/x86/entry/vdso isn't so bad. It's just a bit odd sounding to me.

--Andy

Andy Lutomirski
AMA Capital Management, LLC

We could make it arch/x86/sys/? Sounds a bit too generic though.

Didn't want to limit it to system calls only, because there's various other entry methods (irqs, traps, NMI, etc.) that we want to handle in a coherent fashion. [ Which you are intimately aware of ;-) ]

Thanks,

Ingo

Another tweak would be to move the kernel side entry code into arch/x86/entry/system/ or so, to create the following organization:

arch/x86/entry: all things entry methods

arch/x86/entry/system/: system/kernel mode entry code
arch/x86/entry/vdso/: user mode entry code
arch/x86/entry/vsyscall/: [legacy vsyscall entry code]

arch/x86/entry/syscalls/: build-time syscall table generation code

My primary goal is to have them all close to each other, so that we can have better structure, more coherency and easier overview. The names are negotiable, the concept is not ;-)

Thanks,

Ingo

I think I like the approach in the patches you sent better.

Once this is in, I'll rebase my code movement change and send it, although I probably won't call the result arch/x86/entry/entry.c :) And then it'll be back to grumbling at context tracking.

--Andy

Ok.

Sounds good to me!

A few naming suggestions:

arch/x86/entry/core.c
arch/x86/entry/sys.c
arch/x86/entry/kernel.c

;-)

Note ... I'd love to split up the various entry methods into separate files - but will do that only on top of your C based bits.

And once that is done, the names of those files will be obvious and clean:

arch/x86/entry/syscall.c
arch/x86/entry/irq.c
arch/x86/entry/traps.c
arch/x86/entry/nmi.c
arch/x86/entry/et.c

Uh oh ...

Thanks,

Ingo

I'm convinced. If "entry" means entry into code provided by the kernel as opposed to entry via hardware-provided entry mechanism, then the vdso is indeed a pile of entries into the kernel.

--Andy

Andy Lutomirski
AMA Capital Management, LLC

No, it isn't. It is user space code provided by the kernel. It is a very different beast, and this is highly and unnecessarily confusing. I am thrilled we finally have managed to get the vdso code separated out, this just muddles the situation.

-hpa

So let me clarify this: vdso code is really a kind of system calls, that just happen to be executed in user space, thus *bypassing* kernel entry.
We don't want to put all the system call code (or even the x86-specific
ones) in entry/ as that would completely eliminate the point and we have once again re-invented the kernel/ catch-all.

-hpa

Sort of like this?

https://git.kernel.org/cgit/linux/kernel/git/luto/linux.git/commit/?h=x86/entry&id=d9e37f908e8788c6d50959257f95e279ebdc821d

/me cringes at the impending rebase.

Anyway, I like this series except patch 7.

--Andy

I can't count. I mean all except patch 3 (the vdso one), not 7.

Although arch/x86/entry might be less of a mouthful.

Andy Lutomirski
AMA Capital Management, LLC

So see my reply to hpa: it makes sense to collect all things system calls and other entry code in a single place, instead of having it scattered all around.

Its internal organization is kept intact, so the vDSO code isn't mixed with other bits.

Thanks,

Ingo

Same here... erk.

That really doesn't change the fact that it is a completely different beast, and putting it under entry/ is actively misleading. Having vdso/ as a separate top level is a good thing, I believe.

I really dislike deep hierarchies, as I said, because my experience has been that they are more confusing than they are helpful, especially over time. The arch-example of badness is of course arch/x86/kernel, which is a total catch-all, but

-hpa

Well, directory hierarchy can certainly be overdone, but in general the Linux kernel has the exact opposite problem: too shallow directory organization due to two decades of organic growth.

We've been actively working over the last two years to improve the source code hierarchy gradually: kernel/rcu/, kernel/locking/, kernel/events/, kernel/sched/ and kernel/time/ are all examples of that process. It worked out pretty well IMHO.

This series actively depopulates various catch-all places such as arch/x86/kernel/ and arch/x86/lib/ and moves related code closer to each other.

The purpose of arch/x86/entry/ is:

- implementation of the high level runtime interfaces to enter/exit (call) the Linux kernel, both for hardware and for software defined ABIs.

Each of the files moved directly relate to these interfaces. It's not limited to kernel mode entries alone.

We could rename it to arch/x86/abi/, although that would be somewhat confusing I suspect because it's too generic a name?

Anyway, I pushed this out as it's a step forward and to not hold up ongoing asm work - we'll see how it works out in practice!

Thanks,

Ingo