Skip to content

K
Kernel

openeuler / Kernel
大约 1 年 前同步成功

通知 5
Star 0
Fork 0
K
Kernel

体验新版 GitCode，发现更多精彩内容 >>

切换分支/标签
  1. 12 4月, 2022 2 次提交
  3. 11 4月, 2022 3 次提交
  5. 02 4月, 2022 13 次提交
  7. 30 3月, 2022 3 次提交
  9. 26 3月, 2022 5 次提交
    • Z
      livepatch/arm64: Fix incorrect endian conversion when long jump · 2b491a2e
      由 Zheng Yejian 提交于 
      hulk inclusion
category: bugfix
bugzilla: https://gitee.com/openeuler/kernel/issues/I4ZIB2
CVE: NA

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

Kernel panic happened on 'arm64 big endian' board after calling function
that has been live-patched. It can be reproduced as follows:
  1. Insert 'livepatch-sample.ko' to patch kernel function 'cmdline_proc_show';
  2. Enable patch by execute:
     echo 1 > /sys/kernel/livepatch/livepatch-sample/enabled
  3. Call 'cmdline_proc_show' by execute:
     cat /proc/cmdline
  4. Then we get following panic logs:
     > kernel BUG at arch/arm64/kernel/traps.c:408!
     > Internal error: Oops - BUG: 0 [#1] SMP
     > Modules linked in: dump_mem(OE) livepatch_cmdline1(OEK) hi1382_gmac(OE)
     > [last unloaded: dump_mem]
     > CPU: 3 PID: 1752 Comm: cat Session: 0 Tainted: G           OE K
     > 5.10.0+ #2
     > Hardware name: Hisilicon PhosphorHi1382 (DT)
     > pstate: 00000005 (nzcv daif -PAN -UAO -TCO BTYPE=--)
     > pc : do_undefinstr+0x23c/0x2b4
     > lr : do_undefinstr+0x5c/0x2b4
     > sp : ffffffc010ac3a80
     > x29: ffffffc010ac3a80 x28: ffffff82eb0a8000
     > x27: 0000000000000000 x26: 0000000000000001
     > x25: 0000000000000000 x24: 0000000000001000
     > x23: 0000000000000000 x22: ffffffd0e0f16000
     > x21: ffffffd0e0ae7000 x20: ffffffc010ac3b00
     > x19: 0000000000021fd6 x18: ffffffd0e04aad94
     > x17: 0000000000000000 x16: 0000000000000000
     > x15: ffffffd0e04b519c x14: 0000000000000000
     > x13: 0000000000000000 x12: 0000000000000000
     > x11: 0000000000000000 x10: 0000000000000000
     > x9 : 0000000000000000 x8 : 0000000000000000
     > x7 : 0000000000000000 x6 : ffffffd0e0f16100
     > x5 : 0000000000000000 x4 : 00000000d5300000
     > x3 : 0000000000000000 x2 : ffffffd0e0f160f0
     > x1 : ffffffd0e0f16103 x0 : 0000000000000005
     > Call trace:
     >  do_undefinstr+0x23c/0x2b4
     >  el1_undef+0x2c/0x44
     >  el1_sync_handler+0xa4/0xb0
     >  el1_sync+0x74/0x100
     >  cmdline_proc_show+0xc/0x44
     >  proc_reg_read_iter+0xb0/0xc4
     >  new_sync_read+0x10c/0x15c
     >  vfs_read+0x144/0x18c
     >  ksys_read+0x78/0xe8
     >  __arm64_sys_read+0x24/0x30

We compare first 6 instructions of 'cmdline_proc_show' before and after
patch (see below). There are 4 instructions modified, so this is case
that offset between old and new function is out of 128M. And we found
that instruction at 'cmdline_proc_show+0xc' seems incorrect (it expects
to be '00021fd6').
  origin:     patched:
  --------    --------
  fd7bbea9    929ff7f0
  21d500f0    f2a91b30
  fd030091    f2d00010
  211040f9    d61f0200 <-- cmdline_proc_show+0xc (expect is '00021fd6')
  f30b00f9    f30b00f9
  f30300aa    f30300aa

It is caused by an incorrect big-to-little endian conversion, and we
correct it.

Fixes: 5aa9a1a3 ("livepatch/arm64: support livepatch without ftrace")
Signed-off-by: NZheng Yejian <zhengyejian1@huawei.com>
Reviewed-by: NKuohai Xu <xukuohai@huawei.com>
Signed-off-by: NYongqiang Liu <liuyongqiang13@huawei.com>
      2b491a2e
    • W
      arm64/mpam: realign step entry when traversing rmid_transform · e79aa56b
      由 Wang ShaoBo 提交于 
      hulk inclusion
category: feature
bugzilla: https://gitee.com/openeuler/kernel/issues/I4SE03
CVE: NA

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

This makes step entry aligned with step_size*step_cnt but not step_size,
and check for alignment before traversing rmid_transform.

When modifying rmid with a value not aligned with step_size*step_cnt,
for_each_rmid_transform_point_step_from might miss next step point if
it has been occupied in case step_cnt or step_size not equals to 1,
which will cause the actual allocated rmid to be inconsistent with the
expected one.

Fixes: b47b9a81 ("arm64/mpam: rmid: refine allocation and release process")
Signed-off-by: NWang ShaoBo <bobo.shaobowang@huawei.com>
Reviewed-by: NCheng Jian <cj.chengjian@huawei.com>
Signed-off-by: NLaibin Qiu <qiulaibin@huawei.com>
      e79aa56b
    • X
      dt-bindings: mpam: refactor device tree node structure · 880b1947
      由 Xingang Wang 提交于 
      arm64/mpam: refactor device tree structure to support multiple
devices

ascend inclusion
category: feature
bugzilla:
https://gitee.com/openeuler/kernel/issues/I49RB2
CVE: NA

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

To support multiple mpam device nodes, all nodes should be organized
as child of the same parent nodes. This makes sure that the mpam
discovery start and complete procedure in the right execution order.
Add modification in the devicetree documentation to record this.
Signed-off-by: NXingang Wang <wangxingang5@huawei.com>
Signed-off-by: NWang ShaoBo <bobo.shaobowang@huawei.com>
Reviewed-by: NCheng Jian <cj.chengjian@huawei.com>
Signed-off-by: NLaibin Qiu <qiulaibin@huawei.com>
      880b1947
    • X
      arm64/mpam: refactor device tree structure to support multiple devices · 33dea90e
      由 Xingang Wang 提交于 
      ascend inclusion
category: feature
bugzilla:
https://gitee.com/openeuler/kernel/issues/I49RB2
CVE: NA

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

The process of MPAM device tree initialization is like this:
arm_mpam_device_probe() 	// driver probe
  mpam_discovery_start()	// start discover mpam devices
    [...] 			// find and add mpam devices
  mpam_discovery_complete()   	// trigger mpam_enable

When there are multiple mpam device nodes, the driver probe procedure
will execute more than once. However, the mpam_discovery_start() and
mpam_discovery_complete() should only run once. Besides, the start
should run first, and the complete should run after all devices added.

So we reorganize the device tree structure, so that there will be only
one mpam device parent nodes, and the probe procedure will only run once.
We add the child node to represent the mpam devices, and traverse and
add all mpam devices in the middle procedure of driver probe.
Signed-off-by: NXingang Wang <wangxingang5@huawei.com>
Signed-off-by: NWang ShaoBo <bobo.shaobowang@huawei.com>
Reviewed-by: NCheng Jian <cj.chengjian@huawei.com>
Signed-off-by: NLaibin Qiu <qiulaibin@huawei.com>
      33dea90e
    • X
      arm64/mpam: fix __mpam_device_create() section mismatch error · 11e2bfdd
      由 Xingang Wang 提交于 
      ascend inclusion
category: bugfix
bugzilla: https://gitee.com/openeuler/kernel/issues/I49RB2
CVE: NA

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

Fix modpost Section mismatch error in __mpam_device_create() and others.
These warnings will occur in high version gcc, for example 10.1.0.

  [...]
  WARNING: vmlinux.o(.text+0x2ed88): Section mismatch in reference from the
  function __mpam_device_create() to the function .init.text:mpam_device_alloc()
  The function __mpam_device_create() references
  the function __init mpam_device_alloc().
  This is often because __mpam_device_create lacks a __init
  annotation or the annotation of mpam_device_alloc is wrong.

  WARNING: vmlinux.o(.text.unlikely+0xa5c): Section mismatch in reference from
  the function mpam_resctrl_init() to the function .init.text:mpam_init_padding()
  The function mpam_resctrl_init() references
  the function __init mpam_init_padding().
  This is often because mpam_resctrl_init lacks a __init
  annotation or the annotation of mpam_init_padding is wrong.

  WARNING: vmlinux.o(.text.unlikely+0x5a9c): Section mismatch in reference from
  the function resctrl_group_init() to the function .init.text:resctrl_group_setup_root()
  The function resctrl_group_init() references
  the function __init resctrl_group_setup_root().
  This is often because resctrl_group_init lacks a __init
  annotation or the annotation of resctrl_group_setup_root is wrong.
  [...]

Fixes: c5e27c39 ("arm64/mpam: remove __init macro to support driver probe")
Signed-off-by: NXingang Wang <wangxingang5@huawei.com>
Signed-off-by: NWang ShaoBo <bobo.shaobowang@huawei.com>
Reviewed-by: NCheng Jian <cj.chengjian@huawei.com>
Signed-off-by: NLaibin Qiu <qiulaibin@huawei.com>
      11e2bfdd
  11. 24 3月, 2022 2 次提交
  13. 23 3月, 2022 2 次提交
    • H
      swiotlb: rework "fix info leak with DMA_FROM_DEVICE" · 3f80e186
      由 Halil Pasic 提交于 
      mainline inclusion
from mainline-v5.17-rc8
commit aa6f8dcb
category: bugfix
bugzilla: 186478, https://gitee.com/openeuler/kernel/issues/I4Z86P
CVE: CVE-2022-0854

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

Unfortunately, we ended up merging an old version of the patch "fix info
leak with DMA_FROM_DEVICE" instead of merging the latest one. Christoph
(the swiotlb maintainer), he asked me to create an incremental fix
(after I have pointed this out the mix up, and asked him for guidance).
So here we go.

The main differences between what we got and what was agreed are:
* swiotlb_sync_single_for_device is also required to do an extra bounce
* We decided not to introduce DMA_ATTR_OVERWRITE until we have exploiters
* The implantation of DMA_ATTR_OVERWRITE is flawed: DMA_ATTR_OVERWRITE
  must take precedence over DMA_ATTR_SKIP_CPU_SYNC

Thus this patch removes DMA_ATTR_OVERWRITE, and makes
swiotlb_sync_single_for_device() bounce unconditionally (that is, also
when dir == DMA_TO_DEVICE) in order do avoid synchronising back stale
data from the swiotlb buffer.

Let me note, that if the size used with dma_sync_* API is less than the
size used with dma_[un]map_*, under certain circumstances we may still
end up with swiotlb not being transparent. In that sense, this is no
perfect fix either.

To get this bullet proof, we would have to bounce the entire
mapping/bounce buffer. For that we would have to figure out the starting
address, and the size of the mapping in
swiotlb_sync_single_for_device(). While this does seem possible, there
seems to be no firm consensus on how things are supposed to work.
Signed-off-by: NHalil Pasic <pasic@linux.ibm.com>
Fixes: ddbd89de ("swiotlb: fix info leak with DMA_FROM_DEVICE")
Cc: stable@vger.kernel.org
Reviewed-by: NChristoph Hellwig <hch@lst.de>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>
Conflicts:
	Documentation/core-api/dma-attributes.rst
	include/linux/dma-mapping.h
	kernel/dma/swiotlb.c
Signed-off-by: NLiu Shixin <liushixin2@huawei.com>
Reviewed-by: NXiu Jianfeng <xiujianfeng@huawei.com>
Reviewed-by: NKefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: NYongqiang Liu <liuyongqiang13@huawei.com>
      3f80e186
    • H
      swiotlb: fix info leak with DMA_FROM_DEVICE · 04c20fc8
      由 Halil Pasic 提交于 
      mainline inclusion
from mainline-v5.17-rc6
commit ddbd89de
category: bugfix
bugzilla: 186478, https://gitee.com/openeuler/kernel/issues/I4Z86P
CVE: CVE-2022-0854

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

The problem I'm addressing was discovered by the LTP test covering
cve-2018-1000204.

A short description of what happens follows:
1) The test case issues a command code 00 (TEST UNIT READY) via the SG_IO
   interface with: dxfer_len == 524288, dxdfer_dir == SG_DXFER_FROM_DEV
   and a corresponding dxferp. The peculiar thing about this is that TUR
   is not reading from the device.
2) In sg_start_req() the invocation of blk_rq_map_user() effectively
   bounces the user-space buffer. As if the device was to transfer into
   it. Since commit a45b599a ("scsi: sg: allocate with __GFP_ZERO in
   sg_build_indirect()") we make sure this first bounce buffer is
   allocated with GFP_ZERO.
3) For the rest of the story we keep ignoring that we have a TUR, so the
   device won't touch the buffer we prepare as if the we had a
   DMA_FROM_DEVICE type of situation. My setup uses a virtio-scsi device
   and the  buffer allocated by SG is mapped by the function
   virtqueue_add_split() which uses DMA_FROM_DEVICE for the "in" sgs (here
   scatter-gather and not scsi generics). This mapping involves bouncing
   via the swiotlb (we need swiotlb to do virtio in protected guest like
   s390 Secure Execution, or AMD SEV).
4) When the SCSI TUR is done, we first copy back the content of the second
   (that is swiotlb) bounce buffer (which most likely contains some
   previous IO data), to the first bounce buffer, which contains all
   zeros.  Then we copy back the content of the first bounce buffer to
   the user-space buffer.
5) The test case detects that the buffer, which it zero-initialized,
  ain't all zeros and fails.

One can argue that this is an swiotlb problem, because without swiotlb
we leak all zeros, and the swiotlb should be transparent in a sense that
it does not affect the outcome (if all other participants are well
behaved).

Copying the content of the original buffer into the swiotlb buffer is
the only way I can think of to make swiotlb transparent in such
scenarios. So let's do just that if in doubt, but allow the driver
to tell us that the whole mapped buffer is going to be overwritten,
in which case we can preserve the old behavior and avoid the performance
impact of the extra bounce.
Signed-off-by: NHalil Pasic <pasic@linux.ibm.com>
Signed-off-by: NChristoph Hellwig <hch@lst.de>
Conflicts:
	Documentation/core-api/dma-attributes.rst
	include/linux/dma-mapping.h
	kernel/dma/swiotlb.c
Signed-off-by: NLiu Shixin <liushixin2@huawei.com>
Reviewed-by: NXiu Jianfeng <xiujianfeng@huawei.com>
Reviewed-by: NKefeng Wang <wangkefeng.wang@huawei.com>
Signed-off-by: NYongqiang Liu <liuyongqiang13@huawei.com>
      04c20fc8
  15. 22 3月, 2022 7 次提交
  17. 21 3月, 2022 3 次提交