Skip to content

K
Kernel

openeuler / Kernel
1 年多 前同步成功

通知 8
Star 0
Fork 0
K
Kernel
切换分支/标签
  1. 23 4月, 2022 11 次提交
  3. 16 4月, 2022 1 次提交
    • O
      mm/vmalloc: fix spinning drain_vmap_work after reading from /proc/vmcore · c12cd77c
      由 Omar Sandoval 提交于 
      Commit 3ee48b6a ("mm, x86: Saving vmcore with non-lazy freeing of
vmas") introduced set_iounmap_nonlazy(), which sets vmap_lazy_nr to
lazy_max_pages() + 1, ensuring that any future vunmaps() immediately
purge the vmap areas instead of doing it lazily.

Commit 690467c8 ("mm/vmalloc: Move draining areas out of caller
context") moved the purging from the vunmap() caller to a worker thread.
Unfortunately, set_iounmap_nonlazy() can cause the worker thread to spin
(possibly forever).  For example, consider the following scenario:

 1. Thread reads from /proc/vmcore. This eventually calls
    __copy_oldmem_page() -> set_iounmap_nonlazy(), which sets
    vmap_lazy_nr to lazy_max_pages() + 1.

 2. Then it calls free_vmap_area_noflush() (via iounmap()), which adds 2
    pages (one page plus the guard page) to the purge list and
    vmap_lazy_nr. vmap_lazy_nr is now lazy_max_pages() + 3, so the
    drain_vmap_work is scheduled.

 3. Thread returns from the kernel and is scheduled out.

 4. Worker thread is scheduled in and calls drain_vmap_area_work(). It
    frees the 2 pages on the purge list. vmap_lazy_nr is now
    lazy_max_pages() + 1.

 5. This is still over the threshold, so it tries to purge areas again,
    but doesn't find anything.

 6. Repeat 5.

If the system is running with only one CPU (which is typicial for kdump)
and preemption is disabled, then this will never make forward progress:
there aren't any more pages to purge, so it hangs.  If there is more
than one CPU or preemption is enabled, then the worker thread will spin
forever in the background.  (Note that if there were already pages to be
purged at the time that set_iounmap_nonlazy() was called, this bug is
avoided.)

This can be reproduced with anything that reads from /proc/vmcore
multiple times.  E.g., vmcore-dmesg /proc/vmcore.

It turns out that improvements to vmap() over the years have obsoleted
the need for this "optimization".  I benchmarked `dd if=/proc/vmcore
of=/dev/null` with 4k and 1M read sizes on a system with a 32GB vmcore.
The test was run on 5.17, 5.18-rc1 with a fix that avoided the hang, and
5.18-rc1 with set_iounmap_nonlazy() removed entirely:

    |5.17  |5.18+fix|5.18+removal
  4k|40.86s|  40.09s|      26.73s
  1M|24.47s|  23.98s|      21.84s

The removal was the fastest (by a wide margin with 4k reads).  This
patch removes set_iounmap_nonlazy().

Link: https://lkml.kernel.org/r/52f819991051f9b865e9ce25605509bfdbacadcd.1649277321.git.osandov@fb.com
Fixes: 690467c8  ("mm/vmalloc: Move draining areas out of caller context")
Signed-off-by: NOmar Sandoval <osandov@fb.com>
Acked-by: NChris Down <chris@chrisdown.name>
Reviewed-by: NUladzislau Rezki (Sony) <urezki@gmail.com>
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Acked-by: NBaoquan He <bhe@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
      c12cd77c
  5. 15 4月, 2022 4 次提交
  7. 13 4月, 2022 1 次提交
    • M
      stat: fix inconsistency between struct stat and struct compat_stat · 932aba1e
      由 Mikulas Patocka 提交于 
      struct stat (defined in arch/x86/include/uapi/asm/stat.h) has 32-bit
st_dev and st_rdev; struct compat_stat (defined in
arch/x86/include/asm/compat.h) has 16-bit st_dev and st_rdev followed by
a 16-bit padding.

This patch fixes struct compat_stat to match struct stat.

[ Historical note: the old x86 'struct stat' did have that 16-bit field
  that the compat layer had kept around, but it was changes back in 2003
  by "struct stat - support larger dev_t":

    https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/?id=e95b2065677fe32512a597a79db94b77b90c968d

  and back in those days, the x86_64 port was still new, and separate
  from the i386 code, and had already picked up the old version with a
  16-bit st_dev field ]

Note that we can't change compat_dev_t because it is used by
compat_loop_info.

Also, if the st_dev and st_rdev values are 32-bit, we don't have to use
old_valid_dev to test if the value fits into them.  This fixes
-EOVERFLOW on filesystems that are on NVMe because NVMe uses the major
number 259.
Signed-off-by: NMikulas Patocka <mpatocka@redhat.com>
Cc: Andreas Schwab <schwab@linux-m68k.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
      932aba1e
  9. 12 4月, 2022 4 次提交
  11. 11 4月, 2022 3 次提交
  13. 09 4月, 2022 2 次提交
    • H
      RISC-V: KVM: include missing hwcap.h into vcpu_fp · 4054eee9
      由 Heiko Stuebner 提交于 
      vcpu_fp uses the riscv_isa_extension mechanism which gets
defined in hwcap.h but doesn't include that head file.

While it seems to work in most cases, in certain conditions
this can lead to build failures like

../arch/riscv/kvm/vcpu_fp.c: In function ‘kvm_riscv_vcpu_fp_reset’:
../arch/riscv/kvm/vcpu_fp.c:22:13: error: implicit declaration of function ‘riscv_isa_extension_available’ [-Werror=implicit-function-declaration]
   22 |         if (riscv_isa_extension_available(&isa, f) ||
      |             ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../arch/riscv/kvm/vcpu_fp.c:22:49: error: ‘f’ undeclared (first use in this function)
   22 |         if (riscv_isa_extension_available(&isa, f) ||

Fix this by simply including the necessary header.

Fixes: 0a86512d ("RISC-V: KVM: Factor-out FP virtualization into separate
sources")
Signed-off-by: NHeiko Stuebner <heiko@sntech.de>
Signed-off-by: NAnup Patel <anup@brainfault.org>
      4054eee9
    • A
      RISC-V: KVM: Don't clear hgatp CSR in kvm_arch_vcpu_put() · 8c3ce496
      由 Anup Patel 提交于 
      We might have RISC-V systems (such as QEMU) where VMID is not part
of the TLB entry tag so these systems will have to flush all TLB
entries upon any change in hgatp.VMID.

Currently, we zero-out hgatp CSR in kvm_arch_vcpu_put() and we
re-program hgatp CSR in kvm_arch_vcpu_load(). For above described
systems, this will flush all TLB entries whenever VCPU exits to
user-space hence reducing performance.

This patch fixes above described performance issue by not clearing
hgatp CSR in kvm_arch_vcpu_put().

Fixes: 34bde9d8 ("RISC-V: KVM: Implement VCPU world-switch")
Cc: stable@vger.kernel.org
Signed-off-by: NAnup Patel <apatel@ventanamicro.com>
Signed-off-by: NAnup Patel <anup@brainfault.org>
      8c3ce496
  15. 08 4月, 2022 2 次提交
  17. 07 4月, 2022 12 次提交