linux.git/arch/riscv, branch v5.15 Linux kernel source tree Merge tag 'riscv-for-linus-5.15-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux 2021-10-30T16:28:24+00:00 Linus Torvalds torvalds@linux-foundation.org 2021-10-30T16:28:24+00:00 bf85ba018f9229ce54765a62dba2dea60f7cdafb Pull RISC-V fixes from Palmer Dabbelt: "These are pretty late, but they do fix concrete issues. - ensure the trap vector's address is aligned. - avoid re-populating the KASAN shadow memory. - allow kasan to build without warnings, which have recently become errors" * tag 'riscv-for-linus-5.15-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux: riscv: Fix asan-stack clang build riscv: Do not re-populate shadow memory with kasan_populate_early_shadow riscv: fix misalgned trap vector base address 
Pull RISC-V fixes from Palmer Dabbelt:
 "These are pretty late, but they do fix concrete issues.

   - ensure the trap vector's address is aligned.

   - avoid re-populating the KASAN shadow memory.

   - allow kasan to build without warnings, which have recently become
     errors"

* tag 'riscv-for-linus-5.15-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  riscv: Fix asan-stack clang build
  riscv: Do not re-populate shadow memory with kasan_populate_early_shadow
  riscv: fix misalgned trap vector base address
riscv: Fix asan-stack clang build 2021-10-29T15:54:50+00:00 Alexandre Ghiti alexandre.ghiti@canonical.com 2021-10-29T04:59:27+00:00 54c5639d8f507ebefa814f574cb6f763033a72a5 Nathan reported that because KASAN_SHADOW_OFFSET was not defined in Kconfig, it prevents asan-stack from getting disabled with clang even when CONFIG_KASAN_STACK is disabled: fix this by defining the corresponding config. Reported-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Alexandre Ghiti <alexandre.ghiti@canonical.com> Fixes: 8ad8b72721d0 ("riscv: Add KASAN support") Cc: stable@vger.kernel.org Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com> 
Nathan reported that because KASAN_SHADOW_OFFSET was not defined in
Kconfig, it prevents asan-stack from getting disabled with clang even
when CONFIG_KASAN_STACK is disabled: fix this by defining the
corresponding config.

Reported-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Fixes: 8ad8b72721d0 ("riscv: Add KASAN support")
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
riscv: Do not re-populate shadow memory with kasan_populate_early_shadow 2021-10-29T15:53:42+00:00 Alexandre Ghiti alexandre.ghiti@canonical.com 2021-10-29T04:59:26+00:00 cf11d01135ea1ff7fddb612033e3cb5cde279ff2 When calling this function, all the shadow memory is already populated with kasan_early_shadow_pte which has PAGE_KERNEL protection. kasan_populate_early_shadow write-protects the mapping of the range of addresses passed in argument in zero_pte_populate, which actually write-protects all the shadow memory mapping since kasan_early_shadow_pte is used for all the shadow memory at this point. And then when using memblock API to populate the shadow memory, the first write access to the kernel stack triggers a trap. This becomes visible with the next commit that contains a fix for asan-stack. We already manually populate all the shadow memory in kasan_early_init and we write-protect kasan_early_shadow_pte at the end of kasan_init which makes the calls to kasan_populate_early_shadow superfluous so we can remove them. Signed-off-by: Alexandre Ghiti <alexandre.ghiti@canonical.com> Fixes: e178d670f251 ("riscv/kasan: add KASAN_VMALLOC support") Fixes: 8ad8b72721d0 ("riscv: Add KASAN support") Cc: stable@vger.kernel.org Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com> 
When calling this function, all the shadow memory is already populated
with kasan_early_shadow_pte which has PAGE_KERNEL protection.
kasan_populate_early_shadow write-protects the mapping of the range
of addresses passed in argument in zero_pte_populate, which actually
write-protects all the shadow memory mapping since kasan_early_shadow_pte
is used for all the shadow memory at this point. And then when using
memblock API to populate the shadow memory, the first write access to the
kernel stack triggers a trap. This becomes visible with the next commit
that contains a fix for asan-stack.

We already manually populate all the shadow memory in kasan_early_init
and we write-protect kasan_early_shadow_pte at the end of kasan_init
which makes the calls to kasan_populate_early_shadow superfluous so
we can remove them.

Signed-off-by: Alexandre Ghiti <alexandre.ghiti@canonical.com>
Fixes: e178d670f251 ("riscv/kasan: add KASAN_VMALLOC support")
Fixes: 8ad8b72721d0 ("riscv: Add KASAN support")
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
riscv, bpf: Fix potential NULL dereference 2021-10-28T15:14:30+00:00 Björn Töpel bjorn@kernel.org 2021-10-28T12:51:15+00:00 27de809a3d83a6199664479ebb19712533d6fd9b The bpf_jit_binary_free() function requires a non-NULL argument. When the RISC-V BPF JIT fails to converge in NR_JIT_ITERATIONS steps, jit_data->header will be NULL, which triggers a NULL dereference. Avoid this by checking the argument, prior calling the function. Fixes: ca6cb5447cec ("riscv, bpf: Factor common RISC-V JIT code") Signed-off-by: Björn Töpel <bjorn@kernel.org> Acked-by: Daniel Borkmann <daniel@iogearbox.net> Link: https://lore.kernel.org/r/20211028125115.514587-1-bjorn@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> 
The bpf_jit_binary_free() function requires a non-NULL argument. When
the RISC-V BPF JIT fails to converge in NR_JIT_ITERATIONS steps,
jit_data->header will be NULL, which triggers a NULL
dereference. Avoid this by checking the argument, prior calling the
function.

Fixes: ca6cb5447cec ("riscv, bpf: Factor common RISC-V JIT code")
Signed-off-by: Björn Töpel <bjorn@kernel.org>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/r/20211028125115.514587-1-bjorn@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
riscv: fix misalgned trap vector base address 2021-10-27T20:08:01+00:00 Chen Lu 181250012@smail.nju.edu.cn 2021-10-18T05:22:38+00:00 64a19591a2938b170aa736443d5d3bf4c51e1388 The trap vector marked by label .Lsecondary_park must align on a 4-byte boundary, as the {m,s}tvec is defined to require 4-byte alignment. Signed-off-by: Chen Lu <181250012@smail.nju.edu.cn> Reviewed-by: Anup Patel <anup.patel@wdc.com> Fixes: e011995e826f ("RISC-V: Move relocate and few other functions out of __init") Cc: stable@vger.kernel.org Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com> 
The trap vector marked by label .Lsecondary_park must align on a
4-byte boundary, as the {m,s}tvec is defined to require 4-byte
alignment.

Signed-off-by: Chen Lu <181250012@smail.nju.edu.cn>
Reviewed-by: Anup Patel <anup.patel@wdc.com>
Fixes: e011995e826f ("RISC-V: Move relocate and few other functions out of __init")
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
bpf: Define bpf_jit_alloc_exec_limit for riscv JIT 2021-10-23T00:23:53+00:00 Lorenz Bauer lmb@cloudflare.com 2021-10-14T14:25:51+00:00 8f04db78e4e36a5d4858ce841a3e9cc3d69bde36 Expose the maximum amount of useable memory from the riscv JIT. Signed-off-by: Lorenz Bauer <lmb@cloudflare.com> Signed-off-by: Alexei Starovoitov <ast@kernel.org> Acked-by: Luke Nelson <luke.r.nels@gmail.com> Acked-by: Björn Töpel <bjorn@kernel.org> Link: https://lore.kernel.org/bpf/20211014142554.53120-2-lmb@cloudflare.com 
Expose the maximum amount of useable memory from the riscv JIT.

Signed-off-by: Lorenz Bauer <lmb@cloudflare.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Acked-by: Luke Nelson <luke.r.nels@gmail.com>
Acked-by: Björn Töpel <bjorn@kernel.org>
Link: https://lore.kernel.org/bpf/20211014142554.53120-2-lmb@cloudflare.com
Merge tag 'riscv-for-linus-5.15-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux 2021-10-09T16:07:58+00:00 Linus Torvalds torvalds@linux-foundation.org 2021-10-09T16:07:58+00:00 717478d89fe22df61a4ecf73b1adb31b5f8d1bba Pull RISC-V fixes from Palmer Dabbelt: - A pair of fixes (along with the necessory cleanup) to our VDSO, to avoid a locking during OOM and to prevent the text from overflowing into the data page - A fix to checksyscalls to teach it about our rv32 UABI - A fix to add clone3() to the rv32 UABI, which was pointed out by checksyscalls - A fix to properly flush the icache on the local CPU in addition to the remote CPUs * tag 'riscv-for-linus-5.15-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux: checksyscalls: Unconditionally ignore fstat{,at}64 riscv: Flush current cpu icache before other cpus RISC-V: Include clone3() on rv32 riscv/vdso: make arch_setup_additional_pages wait for mmap_sem for write killable riscv/vdso: Move vdso data page up front riscv/vdso: Refactor asm/vdso.h 
Pull RISC-V fixes from Palmer Dabbelt:

 - A pair of fixes (along with the necessory cleanup) to our VDSO, to
   avoid a locking during OOM and to prevent the text from overflowing
   into the data page

 - A fix to checksyscalls to teach it about our rv32 UABI

 - A fix to add clone3() to the rv32 UABI, which was pointed out by
   checksyscalls

 - A fix to properly flush the icache on the local CPU in addition to
   the remote CPUs

* tag 'riscv-for-linus-5.15-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
  checksyscalls: Unconditionally ignore fstat{,at}64
  riscv: Flush current cpu icache before other cpus
  RISC-V: Include clone3() on rv32
  riscv/vdso: make arch_setup_additional_pages wait for mmap_sem for write killable
  riscv/vdso: Move vdso data page up front
  riscv/vdso: Refactor asm/vdso.h
Merge tag 'asm-generic-fixes-5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic 2021-10-08T18:57:54+00:00 Linus Torvalds torvalds@linux-foundation.org 2021-10-08T18:57:54+00:00 0dcf60d0014001a2dcae61169c8f72b775c2e48a Pull asm-generic fixes from Arnd Bergmann: "There is one build fix for Arm platforms that ended up impacting most architectures because of the way the drivers/firmware Kconfig file is wired up: The CONFIG_QCOM_SCM dependency have caused a number of randconfig regressions over time, and some still remain in v5.15-rc4. The fix we agreed on in the end is to make this symbol selected by any driver using it, and then building it even for non-Arm platforms with CONFIG_COMPILE_TEST. To make this work on all architectures, the drivers/firmware/Kconfig file needs to be included for all architectures to make the symbol itself visible. In a separate discussion, we found that a sound driver patch that is pending for v5.16 needs the same change to include this Kconfig file, so the easiest solution seems to have my Kconfig rework included in v5.15. Finally, the branch also includes a small unrelated build fix for NOMMU architectures" Link: https://lore.kernel.org/all/20210928153508.101208f8@canb.auug.org.au/ Link: https://lore.kernel.org/all/20210928075216.4193128-1-arnd@kernel.org/ Link: https://lore.kernel.org/all/20211007151010.333516-1-arnd@kernel.org/ * tag 'asm-generic-fixes-5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic: asm-generic/io.h: give stub iounmap() on !MMU same prototype as elsewhere qcom_scm: hide Kconfig symbol firmware: include drivers/firmware/Kconfig unconditionally 
Pull asm-generic fixes from Arnd Bergmann:
 "There is one build fix for Arm platforms that ended up impacting most
  architectures because of the way the drivers/firmware Kconfig file is
  wired up:

  The CONFIG_QCOM_SCM dependency have caused a number of randconfig
  regressions over time, and some still remain in v5.15-rc4. The fix we
  agreed on in the end is to make this symbol selected by any driver
  using it, and then building it even for non-Arm platforms with
  CONFIG_COMPILE_TEST.

  To make this work on all architectures, the drivers/firmware/Kconfig
  file needs to be included for all architectures to make the symbol
  itself visible.

  In a separate discussion, we found that a sound driver patch that is
  pending for v5.16 needs the same change to include this Kconfig file,
  so the easiest solution seems to have my Kconfig rework included in
  v5.15.

  Finally, the branch also includes a small unrelated build fix for
  NOMMU architectures"

Link: https://lore.kernel.org/all/20210928153508.101208f8@canb.auug.org.au/
Link: https://lore.kernel.org/all/20210928075216.4193128-1-arnd@kernel.org/
Link: https://lore.kernel.org/all/20211007151010.333516-1-arnd@kernel.org/

* tag 'asm-generic-fixes-5.15' of git://git.kernel.org/pub/scm/linux/kernel/git/arnd/asm-generic:
  asm-generic/io.h: give stub iounmap() on !MMU same prototype as elsewhere
  qcom_scm: hide Kconfig symbol
  firmware: include drivers/firmware/Kconfig unconditionally
firmware: include drivers/firmware/Kconfig unconditionally 2021-10-07T14:51:26+00:00 Arnd Bergmann arnd@arndb.de 2021-09-28T07:50:26+00:00 951cd3a0866d29cb9c01ebc1d9c17590e598226e Compile-testing drivers that require access to a firmware layer fails when that firmware symbol is unavailable. This happened twice this week: - My proposed to change to rework the QCOM_SCM firmware symbol broke on ppc64 and others. - The cs_dsp firmware patch added device specific firmware loader into drivers/firmware, which broke on the same set of architectures. We should probably do the same thing for other subsystems as well, but fix this one first as this is a dependency for other patches getting merged. Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org> Reviewed-by: Charles Keepax <ckeepax@opensource.cirrus.com> Acked-by: Will Deacon <will@kernel.org> Acked-by: Bjorn Andersson <bjorn.andersson@linaro.org> Cc: Mark Brown <broonie@kernel.org> Cc: Liam Girdwood <lgirdwood@gmail.com> Cc: Charles Keepax <ckeepax@opensource.cirrus.com> Cc: Simon Trimmer <simont@opensource.cirrus.com> Cc: Michael Ellerman <mpe@ellerman.id.au> Reviewed-by: Mark Brown <broonie@kernel.org> Signed-off-by: Arnd Bergmann <arnd@arndb.de> 
Compile-testing drivers that require access to a firmware layer
fails when that firmware symbol is unavailable. This happened
twice this week:

 - My proposed to change to rework the QCOM_SCM firmware symbol
   broke on ppc64 and others.

 - The cs_dsp firmware patch added device specific firmware loader
   into drivers/firmware, which broke on the same set of
   architectures.

We should probably do the same thing for other subsystems as well,
but fix this one first as this is a dependency for other patches
getting merged.

Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Reviewed-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Acked-by: Will Deacon <will@kernel.org>
Acked-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Cc: Mark Brown <broonie@kernel.org>
Cc: Liam Girdwood <lgirdwood@gmail.com>
Cc: Charles Keepax <ckeepax@opensource.cirrus.com>
Cc: Simon Trimmer <simont@opensource.cirrus.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Reviewed-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
riscv: Flush current cpu icache before other cpus 2021-10-05T01:24:15+00:00 Alexandre Ghiti alex@ghiti.fr 2021-09-18T16:02:21+00:00 bb8958d5dc79acbd071397abb57b8756375fe1ce On SiFive Unmatched, I recently fell onto the following BUG when booting: [ 0.000000] ftrace: allocating 36610 entries in 144 pages [ 0.000000] Oops - illegal instruction [#1] [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 5.13.1+ #5 [ 0.000000] Hardware name: SiFive HiFive Unmatched A00 (DT) [ 0.000000] epc : riscv_cpuid_to_hartid_mask+0x6/0xae [ 0.000000] ra : __sbi_rfence_v02+0xc8/0x10a [ 0.000000] epc : ffffffff80007240 ra : ffffffff80009964 sp : ffffffff81803e10 [ 0.000000] gp : ffffffff81a1ea70 tp : ffffffff8180f500 t0 : ffffffe07fe30000 [ 0.000000] t1 : 0000000000000004 t2 : 0000000000000000 s0 : ffffffff81803e60 [ 0.000000] s1 : 0000000000000000 a0 : ffffffff81a22238 a1 : ffffffff81803e10 [ 0.000000] a2 : 0000000000000000 a3 : 0000000000000000 a4 : 0000000000000000 [ 0.000000] a5 : 0000000000000000 a6 : ffffffff8000989c a7 : 0000000052464e43 [ 0.000000] s2 : ffffffff81a220c8 s3 : 0000000000000000 s4 : 0000000000000000 [ 0.000000] s5 : 0000000000000000 s6 : 0000000200000100 s7 : 0000000000000001 [ 0.000000] s8 : ffffffe07fe04040 s9 : ffffffff81a22c80 s10: 0000000000001000 [ 0.000000] s11: 0000000000000004 t3 : 0000000000000001 t4 : 0000000000000008 [ 0.000000] t5 : ffffffcf04000808 t6 : ffffffe3ffddf188 [ 0.000000] status: 0000000200000100 badaddr: 0000000000000000 cause: 0000000000000002 [ 0.000000] [<ffffffff80007240>] riscv_cpuid_to_hartid_mask+0x6/0xae [ 0.000000] [<ffffffff80009474>] sbi_remote_fence_i+0x1e/0x26 [ 0.000000] [<ffffffff8000b8f4>] flush_icache_all+0x12/0x1a [ 0.000000] [<ffffffff8000666c>] patch_text_nosync+0x26/0x32 [ 0.000000] [<ffffffff8000884e>] ftrace_init_nop+0x52/0x8c [ 0.000000] [<ffffffff800f051e>] ftrace_process_locs.isra.0+0x29c/0x360 [ 0.000000] [<ffffffff80a0e3c6>] ftrace_init+0x80/0x130 [ 0.000000] [<ffffffff80a00f8c>] start_kernel+0x5c4/0x8f6 [ 0.000000] ---[ end trace f67eb9af4d8d492b ]--- [ 0.000000] Kernel panic - not syncing: Attempted to kill the idle task! [ 0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]--- While ftrace is looping over a list of addresses to patch, it always failed when patching the same function: riscv_cpuid_to_hartid_mask. Looking at the backtrace, the illegal instruction is encountered in this same function. However, patch_text_nosync, after patching the instructions, calls flush_icache_range. But looking at what happens in this function: flush_icache_range -> flush_icache_all -> sbi_remote_fence_i -> __sbi_rfence_v02 -> riscv_cpuid_to_hartid_mask The icache and dcache of the current cpu are never synchronized between the patching of riscv_cpuid_to_hartid_mask and calling this same function. So fix this by flushing the current cpu's icache before asking for the other cpus to do the same. Signed-off-by: Alexandre Ghiti <alex@ghiti.fr> Fixes: fab957c11efe ("RISC-V: Atomic and Locking Code") Cc: stable@vger.kernel.org Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com> 
On SiFive Unmatched, I recently fell onto the following BUG when booting:

[    0.000000] ftrace: allocating 36610 entries in 144 pages
[    0.000000] Oops - illegal instruction [#1]
[    0.000000] Modules linked in:
[    0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 5.13.1+ #5
[    0.000000] Hardware name: SiFive HiFive Unmatched A00 (DT)
[    0.000000] epc : riscv_cpuid_to_hartid_mask+0x6/0xae
[    0.000000]  ra : __sbi_rfence_v02+0xc8/0x10a
[    0.000000] epc : ffffffff80007240 ra : ffffffff80009964 sp : ffffffff81803e10
[    0.000000]  gp : ffffffff81a1ea70 tp : ffffffff8180f500 t0 : ffffffe07fe30000
[    0.000000]  t1 : 0000000000000004 t2 : 0000000000000000 s0 : ffffffff81803e60
[    0.000000]  s1 : 0000000000000000 a0 : ffffffff81a22238 a1 : ffffffff81803e10
[    0.000000]  a2 : 0000000000000000 a3 : 0000000000000000 a4 : 0000000000000000
[    0.000000]  a5 : 0000000000000000 a6 : ffffffff8000989c a7 : 0000000052464e43
[    0.000000]  s2 : ffffffff81a220c8 s3 : 0000000000000000 s4 : 0000000000000000
[    0.000000]  s5 : 0000000000000000 s6 : 0000000200000100 s7 : 0000000000000001
[    0.000000]  s8 : ffffffe07fe04040 s9 : ffffffff81a22c80 s10: 0000000000001000
[    0.000000]  s11: 0000000000000004 t3 : 0000000000000001 t4 : 0000000000000008
[    0.000000]  t5 : ffffffcf04000808 t6 : ffffffe3ffddf188
[    0.000000] status: 0000000200000100 badaddr: 0000000000000000 cause: 0000000000000002
[    0.000000] [<ffffffff80007240>] riscv_cpuid_to_hartid_mask+0x6/0xae
[    0.000000] [<ffffffff80009474>] sbi_remote_fence_i+0x1e/0x26
[    0.000000] [<ffffffff8000b8f4>] flush_icache_all+0x12/0x1a
[    0.000000] [<ffffffff8000666c>] patch_text_nosync+0x26/0x32
[    0.000000] [<ffffffff8000884e>] ftrace_init_nop+0x52/0x8c
[    0.000000] [<ffffffff800f051e>] ftrace_process_locs.isra.0+0x29c/0x360
[    0.000000] [<ffffffff80a0e3c6>] ftrace_init+0x80/0x130
[    0.000000] [<ffffffff80a00f8c>] start_kernel+0x5c4/0x8f6
[    0.000000] ---[ end trace f67eb9af4d8d492b ]---
[    0.000000] Kernel panic - not syncing: Attempted to kill the idle task!
[    0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---

While ftrace is looping over a list of addresses to patch, it always failed
when patching the same function: riscv_cpuid_to_hartid_mask. Looking at the
backtrace, the illegal instruction is encountered in this same function.
However, patch_text_nosync, after patching the instructions, calls
flush_icache_range. But looking at what happens in this function:

flush_icache_range -> flush_icache_all
                   -> sbi_remote_fence_i
                   -> __sbi_rfence_v02
                   -> riscv_cpuid_to_hartid_mask

The icache and dcache of the current cpu are never synchronized between the
patching of riscv_cpuid_to_hartid_mask and calling this same function.

So fix this by flushing the current cpu's icache before asking for the other
cpus to do the same.

Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Fixes: fab957c11efe ("RISC-V: Atomic and Locking Code")
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>