https://github.com/torvalds/linux
Revision 3ad33b2436b545cbe8b28e53f3710432cad457ab authored by Lee Schermerhorn on 15 November 2007, 00:59:10 UTC, committed by Linus Torvalds on 15 November 2007, 02:45:38 UTC
We hit the BUG_ON() in mm/rmap.c:vma_address() when trying to migrate via
mbind(MPOL_MF_MOVE) a non-anon region that spans multiple vmas.  For
anon-regions, we just fail to migrate any pages beyond the 1st vma in the
range.

This occurs because do_mbind() collects a list of pages to migrate by
calling check_range().  check_range() walks the task's mm, spanning vmas as
necessary, to collect the migratable pages into a list.  Then, do_mbind()
calls migrate_pages() passing the list of pages, a function to allocate new
pages based on vma policy [new_vma_page()], and a pointer to the first vma
of the range.

For each page in the list, new_vma_page() calls page_address_in_vma()
passing the page and the vma [first in range] to obtain the address to get
for alloc_page_vma().  The page address is needed to get interleaving
policy correct.  If the pages in the list come from multiple vmas,
eventually, new_page_address() will pass that page to page_address_in_vma()
with the incorrect vma.  For !PageAnon pages, this will result in a bug
check in rmap.c:vma_address().  For anon pages, vma_address() will just
return EFAULT and fail the migration.

This patch modifies new_vma_page() to check the return value from
page_address_in_vma().  If the return value is EFAULT, new_vma_page()
searchs forward via vm_next for the vma that maps the page--i.e., that does
not return EFAULT.  This assumes that the pages in the list handed to
migrate_pages() is in address order.  This is currently case.  The patch
documents this assumption in a new comment block for new_vma_page().

If new_vma_page() cannot locate the vma mapping the page in a forward
search in the mm, it will pass a NULL vma to alloc_page_vma().  This will
result in the allocation using the task policy, if any, else system default
policy.  This situation is unlikely, but the patch documents this behavior
with a comment.

Note, this patch results in restarting from the first vma in a multi-vma
range each time new_vma_page() is called.  If this is not acceptable, we
can make the vma argument a pointer, both in new_vma_page() and it's caller
unmap_and_move() so that the value held by the loop in migrate_pages()
always passes down the last vma in which a page was found.  This will
require changes to all new_page_t functions passed to migrate_pages().  Is
this necessary?

For this patch to work, we can't bug check in vma_address() for pages
outside the argument vma.  This patch removes the BUG_ON().  All other
callers [besides new_vma_page()] already check the return status.

Tested on x86_64, 4 node NUMA platform.

Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
1 parent e1a1c99
Raw File
Tip revision: 3ad33b2436b545cbe8b28e53f3710432cad457ab authored by Lee Schermerhorn on 15 November 2007, 00:59:10 UTC
Migration: find correct vma in new_vma_page()
Tip revision: 3ad33b2
Makefile.host
# ==========================================================================
# Building binaries on the host system
# Binaries are used during the compilation of the kernel, for example
# to preprocess a data file.
#
# Both C and C++ is supported, but preferred language is C for such utilities.
#
# Samle syntax (see Documentation/kbuild/makefile.txt for reference)
# hostprogs-y := bin2hex
# Will compile bin2hex.c and create an executable named bin2hex
#
# hostprogs-y    := lxdialog
# lxdialog-objs := checklist.o lxdialog.o
# Will compile lxdialog.c and checklist.c, and then link the executable
# lxdialog, based on checklist.o and lxdialog.o
#
# hostprogs-y      := qconf
# qconf-cxxobjs   := qconf.o
# qconf-objs      := menu.o
# Will compile qconf as a C++ program, and menu as a C program.
# They are linked as C++ code to the executable qconf

# hostprogs-y := conf
# conf-objs  := conf.o libkconfig.so
# libkconfig-objs := expr.o type.o
# Will create a shared library named libkconfig.so that consist of
# expr.o and type.o (they are both compiled as C code and the object file
# are made as position independent code).
# conf.c is compiled as a c program, and conf.o is linked together with
# libkconfig.so as the executable conf.
# Note: Shared libraries consisting of C++ files are not supported

__hostprogs := $(sort $(hostprogs-y) $(hostprogs-m))

# C code
# Executables compiled from a single .c file
host-csingle	:= $(foreach m,$(__hostprogs),$(if $($(m)-objs),,$(m)))

# C executables linked based on several .o files
host-cmulti	:= $(foreach m,$(__hostprogs),\
		   $(if $($(m)-cxxobjs),,$(if $($(m)-objs),$(m))))

# Object (.o) files compiled from .c files
host-cobjs	:= $(sort $(foreach m,$(__hostprogs),$($(m)-objs)))

# C++ code
# C++ executables compiled from at least on .cc file
# and zero or more .c files
host-cxxmulti	:= $(foreach m,$(__hostprogs),$(if $($(m)-cxxobjs),$(m)))

# C++ Object (.o) files compiled from .cc files
host-cxxobjs	:= $(sort $(foreach m,$(host-cxxmulti),$($(m)-cxxobjs)))

# Shared libaries (only .c supported)
# Shared libraries (.so) - all .so files referenced in "xxx-objs"
host-cshlib	:= $(sort $(filter %.so, $(host-cobjs)))
# Remove .so files from "xxx-objs"
host-cobjs	:= $(filter-out %.so,$(host-cobjs))

#Object (.o) files used by the shared libaries
host-cshobjs	:= $(sort $(foreach m,$(host-cshlib),$($(m:.so=-objs))))

# output directory for programs/.o files
# hostprogs-y := tools/build may have been specified. Retreive directory
host-objdirs := $(foreach f,$(__hostprogs), $(if $(dir $(f)),$(dir $(f))))
# directory of .o files from prog-objs notation
host-objdirs += $(foreach f,$(host-cmulti),                  \
                    $(foreach m,$($(f)-objs),                \
                        $(if $(dir $(m)),$(dir $(m)))))
# directory of .o files from prog-cxxobjs notation
host-objdirs += $(foreach f,$(host-cxxmulti),                  \
                    $(foreach m,$($(f)-cxxobjs),                \
                        $(if $(dir $(m)),$(dir $(m)))))

host-objdirs := $(strip $(sort $(filter-out ./,$(host-objdirs))))


__hostprogs     := $(addprefix $(obj)/,$(__hostprogs))
host-csingle	:= $(addprefix $(obj)/,$(host-csingle))
host-cmulti	:= $(addprefix $(obj)/,$(host-cmulti))
host-cobjs	:= $(addprefix $(obj)/,$(host-cobjs))
host-cxxmulti	:= $(addprefix $(obj)/,$(host-cxxmulti))
host-cxxobjs	:= $(addprefix $(obj)/,$(host-cxxobjs))
host-cshlib	:= $(addprefix $(obj)/,$(host-cshlib))
host-cshobjs	:= $(addprefix $(obj)/,$(host-cshobjs))
host-objdirs    := $(addprefix $(obj)/,$(host-objdirs))

obj-dirs += $(host-objdirs)

#####
# Handle options to gcc. Support building with separate output directory

_hostc_flags   = $(HOSTCFLAGS)   $(HOST_EXTRACFLAGS)   \
                 $(HOSTCFLAGS_$(basetarget).o)
_hostcxx_flags = $(HOSTCXXFLAGS) $(HOST_EXTRACXXFLAGS) \
                 $(HOSTCXXFLAGS_$(basetarget).o)

ifeq ($(KBUILD_SRC),)
__hostc_flags	= $(_hostc_flags)
__hostcxx_flags	= $(_hostcxx_flags)
else
__hostc_flags	= -I$(obj) $(call flags,_hostc_flags)
__hostcxx_flags	= -I$(obj) $(call flags,_hostcxx_flags)
endif

hostc_flags    = -Wp,-MD,$(depfile) $(__hostc_flags)
hostcxx_flags  = -Wp,-MD,$(depfile) $(__hostcxx_flags)

#####
# Compile programs on the host

# Create executable from a single .c file
# host-csingle -> Executable
quiet_cmd_host-csingle 	= HOSTCC  $@
      cmd_host-csingle	= $(HOSTCC) $(hostc_flags) -o $@ $< \
	  	$(HOST_LOADLIBES) $(HOSTLOADLIBES_$(@F))
$(host-csingle): $(obj)/%: $(src)/%.c FORCE
	$(call if_changed_dep,host-csingle)

# Link an executable based on list of .o files, all plain c
# host-cmulti -> executable
quiet_cmd_host-cmulti	= HOSTLD  $@
      cmd_host-cmulti	= $(HOSTCC) $(HOSTLDFLAGS) -o $@ \
			  $(addprefix $(obj)/,$($(@F)-objs)) \
			  $(HOST_LOADLIBES) $(HOSTLOADLIBES_$(@F))
$(host-cmulti): $(obj)/%: $(host-cobjs) $(host-cshlib) FORCE
	$(call if_changed,host-cmulti)

# Create .o file from a single .c file
# host-cobjs -> .o
quiet_cmd_host-cobjs	= HOSTCC  $@
      cmd_host-cobjs	= $(HOSTCC) $(hostc_flags) -c -o $@ $<
$(host-cobjs): $(obj)/%.o: $(src)/%.c FORCE
	$(call if_changed_dep,host-cobjs)

# Link an executable based on list of .o files, a mixture of .c and .cc
# host-cxxmulti -> executable
quiet_cmd_host-cxxmulti	= HOSTLD  $@
      cmd_host-cxxmulti	= $(HOSTCXX) $(HOSTLDFLAGS) -o $@ \
			  $(foreach o,objs cxxobjs,\
			  $(addprefix $(obj)/,$($(@F)-$(o)))) \
			  $(HOST_LOADLIBES) $(HOSTLOADLIBES_$(@F))
$(host-cxxmulti): $(obj)/%: $(host-cobjs) $(host-cxxobjs) $(host-cshlib) FORCE
	$(call if_changed,host-cxxmulti)

# Create .o file from a single .cc (C++) file
quiet_cmd_host-cxxobjs	= HOSTCXX $@
      cmd_host-cxxobjs	= $(HOSTCXX) $(hostcxx_flags) -c -o $@ $<
$(host-cxxobjs): $(obj)/%.o: $(src)/%.cc FORCE
	$(call if_changed_dep,host-cxxobjs)

# Compile .c file, create position independent .o file
# host-cshobjs -> .o
quiet_cmd_host-cshobjs	= HOSTCC  -fPIC $@
      cmd_host-cshobjs	= $(HOSTCC) $(hostc_flags) -fPIC -c -o $@ $<
$(host-cshobjs): $(obj)/%.o: $(src)/%.c FORCE
	$(call if_changed_dep,host-cshobjs)

# Link a shared library, based on position independent .o files
# *.o -> .so shared library (host-cshlib)
quiet_cmd_host-cshlib	= HOSTLLD -shared $@
      cmd_host-cshlib	= $(HOSTCC) $(HOSTLDFLAGS) -shared -o $@ \
			  $(addprefix $(obj)/,$($(@F:.so=-objs))) \
			  $(HOST_LOADLIBES) $(HOSTLOADLIBES_$(@F))
$(host-cshlib): $(obj)/%: $(host-cshobjs) FORCE
	$(call if_changed,host-cshlib)

targets += $(host-csingle)  $(host-cmulti) $(host-cobjs)\
	   $(host-cxxmulti) $(host-cxxobjs) $(host-cshlib) $(host-cshobjs) 

back to top