4.4.26-stable review

This is the start of the stable review cycle for the 4.4.26 release. There are 2 patches in this series, all will be posted as a response to this one. If anyone has any issues with these being applied, please let me know.

Responses should be made by Fri Oct 21 18:27:53 UTC 2016. Anything received after that time might be too late.

The whole patch series can be found in one patch at:

	kernel.org/pub/linux/kernel/v4.x/stable-review/patch-4.4.26-rc1.gz

or in the git tree and branch at:
git://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable-rc.git linux-4.4.y
and the diffstat can be found below.

thanks,

greg k-h

-------------
Pseudo-Shortlog of commits:

Greg Kroah-Hartman <[email protected]>

    Linux 4.4.26-rc1

Linus Torvalds <[email protected]>
mm: remove gup_flags FOLL_WRITE games from __get_user_pages()

H.J. Lu <[email protected]>
x86/build: Build compressed x86 kernels as PIE

-------------

Diffstat:

 Makefile                           |  4 ++--
 arch/x86/boot/compressed/Makefile  | 14 +++++++++++++-
 arch/x86/boot/compressed/head_32.S | 28 ++++++++++++++++++++++++++++
 arch/x86/boot/compressed/head_64.S |  8 ++++++++
 include/linux/mm.h                 |  1 +
 mm/gup.c                           | 14 ++++++++++++--
 6 files changed, 64 insertions(+), 5 deletions(-)

4.4-stable review patch. If anyone has any objections, please let me know.

------------------

From: H.J. Lu <[email protected]>

commit 6d92bc9d483aa1751755a66fee8fb39dffb088c0 upstream.

The 32-bit x86 assembler in binutils 2.26 will generate R_386_GOT32X relocation to get the symbol address in PIC. When the compressed x86 kernel isn't built as PIC, the linker optimizes R_386_GOT32X relocations to their fixed symbol addresses. However, when the compressed x86 kernel is loaded at a different address, it leads to the following load failure:

Failed to allocate space for phdrs

during the decompression stage.

If the compressed x86 kernel is relocatable at run-time, it should be compiled with -fPIE, instead of -fPIC, if possible and should be built as Position Independent Executable (PIE) so that linker won't optimize R_386_GOT32X relocation to its fixed symbol address.

Older linkers generate R_386_32 relocations against locally defined symbols, _bss, _ebss, _got and _egot, in PIE. It isn't wrong, just less optimal than R_386_RELATIVE. But the x86 kernel fails to properly handle R_386_32 relocations when relocating the kernel. To generate R_386_RELATIVE relocations, we mark _bss, _ebss, _got and _egot as hidden in both 32-bit and 64-bit x86 kernels.

To build a 64-bit compressed x86 kernel as PIE, we need to disable the relocation overflow check to avoid relocation overflow errors. We do this with a new linker command-line option, -z noreloc-overflow, which got added recently:

commit 4c10bbaa0912742322f10d9d5bb630ba4e15dfa7
Author: H.J. Lu <[email protected]>
Date: Tue Mar 15 11:07:06 2016 -0700

Add -z noreloc-overflow option to x86-64 ld

Add -z noreloc-overflow command-line option to the x86-64 ELF linker to disable relocation overflow check. This can be used to avoid relocation overflow check if there will be no dynamic relocation overflow at run-time.

The 64-bit compressed x86 kernel is built as PIE only if the linker supports -z noreloc-overflow. So far 64-bit relocatable compressed x86 kernel boots fine even when it is built as a normal executable.

---
 arch/x86/boot/compressed/Makefile  |   14 +++++++++++++-
 arch/x86/boot/compressed/head_32.S |   28 ++++++++++++++++++++++++++++
 arch/x86/boot/compressed/head_64.S |    8 ++++++++
 3 files changed, 49 insertions(+), 1 deletion(-)

@@ -22,7 +22,7 @@ targets := vmlinux vmlinux.bin vmlinux.b
 	vmlinux.bin.xz vmlinux.bin.lzo vmlinux.bin.lz4
 
 KBUILD_CFLAGS := -m$(BITS) -D__KERNEL__ $(LINUX_INCLUDE) -O2
-KBUILD_CFLAGS += -fno-strict-aliasing -fPIC
+KBUILD_CFLAGS += -fno-strict-aliasing $(call cc-option, -fPIE, -fPIC)
 KBUILD_CFLAGS += -DDISABLE_BRANCH_PROFILING
 cflags-$(CONFIG_X86_32) := -march=i386
 cflags-$(CONFIG_X86_64) := -mcmodel=small

@@ -35,6 +35,18 @@ KBUILD_AFLAGS  := $(KBUILD_CFLAGS) -D__A
 GCOV_PROFILE := n
 
 LDFLAGS := -m elf_$(UTS_MACHINE)
+ifeq ($(CONFIG_RELOCATABLE),y)
+# If kernel is relocatable, build compressed kernel as PIE.
+ifeq ($(CONFIG_X86_32),y)
+LDFLAGS += $(call ld-option, -pie) $(call ld-option, --no-dynamic-linker)
+else
+# To build 64-bit compressed kernel as PIE, we disable relocation
+# overflow check to avoid relocation overflow error with a new linker
+# command-line option, -z noreloc-overflow.
+LDFLAGS += $(shell $(LD) --help 2>&1 | grep -q "\-z noreloc-overflow" \
+	&& echo "-z noreloc-overflow -pie --no-dynamic-linker")
+endif
+endif
 LDFLAGS_vmlinux := -T
 
 hostprogs-y	:= mkpiggy

@@ -31,6 +31,34 @@
 #include <asm/asm-offsets.h>
 #include <asm/bootparam.h>
 
+/*
+ * The 32-bit x86 assembler in binutils 2.26 will generate R_386_GOT32X
+ * relocation to get the symbol address in PIC.  When the compressed x86
+ * kernel isn't built as PIC, the linker optimizes R_386_GOT32X
+ * relocations to their fixed symbol addresses.  However, when the
+ * compressed x86 kernel is loaded at a different address, it leads
+ * to the following load failure:
+ *
+ *   Failed to allocate space for phdrs
+ *
+ * during the decompression stage.
+ *
+ * If the compressed x86 kernel is relocatable at run-time, it should be
+ * compiled with -fPIE, instead of -fPIC, if possible and should be built as
+ * Position Independent Executable (PIE) so that linker won't optimize
+ * R_386_GOT32X relocation to its fixed symbol address.  Older
+ * linkers generate R_386_32 relocations against locally defined symbols,
+ * _bss, _ebss, _got and _egot, in PIE.  It isn't wrong, just less
+ * optimal than R_386_RELATIVE.  But the x86 kernel fails to properly handle
+ * R_386_32 relocations when relocating the kernel.  To generate
+ * R_386_RELATIVE relocations, we mark _bss, _ebss, _got and _egot as
+ * hidden:
+ */
+	.hidden _bss
+	.hidden _ebss
+	.hidden _got
+	.hidden _egot
+
 	__HEAD
 ENTRY(startup_32)
 #ifdef CONFIG_EFI_STUB

@@ -33,6 +33,14 @@
 #include <asm/asm-offsets.h>
 #include <asm/bootparam.h>
 
+/*
+ * Locally defined symbols should be marked hidden:
+ */
+	.hidden _bss
+	.hidden _ebss
+	.hidden _got
+	.hidden _egot
+
 	__HEAD
 	.code32
 ENTRY(startup_32)

4.4-stable review patch. If anyone has any objections, please let me know.

------------------

From: Linus Torvalds <[email protected]>

commit 19be0eaffa3ac7d8eb6784ad9bdbc7d67ed8e619 upstream.

This is an ancient bug that was actually attempted to be fixed once (badly) by me eleven years ago in commit 4ceb5db9757a ("Fix get_user_pages() race for write access") but that was then undone due to problems on s390 by commit f33ea7f404e5 ("fix get_user_pages bug").

In the meantime, the s390 situation has long been fixed, and we can now fix it by checking the pte_dirty() bit properly (and do it better). The s390 dirty bit was implemented in abf09bed3cce ("s390/mm: implement software dirty bits") which made it into v3.9. Earlier kernels will have to look at the page state itself.

Also, the VM has become more scalable, and what used a purely theoretical race back then has become easier to trigger.

To fix it, we introduce a new internal FOLL_COW flag to mark the "yes, we already did a COW" rather than play racy games with FOLL_WRITE that is very fundamental, and then use the pte dirty flag to validate that the FOLL_COW flag is still valid.

---
 include/linux/mm.h |    1 +
 mm/gup.c           |   14 ++++++++++++--
 2 files changed, 13 insertions(+), 2 deletions(-)

@@ -2112,6 +2112,7 @@ static inline struct page *follow_page(s
 #define FOLL_MIGRATION	0x400	/* wait for page to replace migration entry */
 #define FOLL_TRIED	0x800	/* a retry, previous pass started an IO */
 #define FOLL_MLOCK	0x1000	/* lock present pages */
+#define FOLL_COW	0x4000	/* internal GUP flag */
 
 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
 			void *data);

@@ -58,6 +58,16 @@ static int follow_pfn_pte(struct vm_area
 	return -EEXIST;
 }
 
+/*
+ * FOLL_FORCE can write to even unwritable pte's, but only
+ * after we've gone through a COW cycle and they are dirty.
+ */
+static inline bool can_follow_write_pte(pte_t pte, unsigned int flags)
+{
+	return pte_write(pte) ||
+		((flags & FOLL_FORCE) && (flags & FOLL_COW) && pte_dirty(pte));
+}
+
 static struct page *follow_page_pte(struct vm_area_struct *vma,
 		unsigned long address, pmd_t *pmd, unsigned int flags)
 {

@@ -92,7 +102,7 @@ retry:
 	}
 	if ((flags & FOLL_NUMA) && pte_protnone(pte))
 		goto no_page;
-	if ((flags & FOLL_WRITE) && !pte_write(pte)) {
+	if ((flags & FOLL_WRITE) && !can_follow_write_pte(pte, flags)) {
 		pte_unmap_unlock(ptep, ptl);
 		return NULL;
 	}

@@ -352,7 +362,7 @@ static int faultin_page(struct task_stru
 	 * reCOWed by userspace write).
 	 */
 	if ((ret & VM_FAULT_WRITE) && !(vma->vm_flags & VM_WRITE))
-		*flags &= ~FOLL_WRITE;
+	        *flags |= FOLL_COW;
 	return 0;
 }
 

Did I botch my attempt at a backport of "lightnvm: ensure that nvm_dev_ops can be used without CONFIG_NVM" to v4.4.y (see
https://lkml.kernel.org/r/<1476477349-28155-1-git-send-email-pebolle@ti
scali.nl> ) sufficiently for it to be dropped?

Thanks,

Paul Bolle

It's in good company, sitting along with 250+ other patches I have yet to work through to apply to the stable kernels. For various reasons I needed to get a round of stable kernels out sooner, which is why it isn't in there. Don't worry, it's not lost, it will get handled eventually...

thanks,

greg k-h

Great. I'll be patient from now on. Sorry for the noise.

Paul Bolle

Compiled and booted on my test system. No dmesg regressions.

thanks,
-- Shuah

Shuah Khan
Sr. Linux Kernel Developer
Open Source Innovation Group
Samsung Research America(Silicon Valley)
[email protected]

Build results:

	total: 149 pass: 149 fail: 0
Qemu test results:
	total: 103 pass: 103 fail: 0

Details are available at http://kerneltests.org/builders.

Guenter