cxl/region: Fix cxl_region_rwsem lock held when returning to user space
authorLi Zhijian <lizhijian@fujitsu.com>
Wed, 25 Oct 2023 08:54:50 +0000 (16:54 +0800)
committerDan Williams <dan.j.williams@intel.com>
Fri, 27 Oct 2023 20:04:51 +0000 (13:04 -0700)
commit3531b27f1f04a6bc9c95cf00d40efe618d57aa93
treeeb8366b6567aa9143e73a6989817200bd5af5b31
parent0cf36a85c1408f86a967fb1db721de1b89b9e675
cxl/region: Fix cxl_region_rwsem lock held when returning to user space

Fix a missed "goto out" to unlock on error to cleanup this splat:

    WARNING: lock held when returning to user space!
    6.6.0-rc3-lizhijian+ #213 Not tainted
    ------------------------------------------------
    cxl/673 is leaving the kernel with locks still held!
    1 lock held by cxl/673:
     #0: ffffffffa013b9d0 (cxl_region_rwsem){++++}-{3:3}, at: commit_store+0x7d/0x3e0 [cxl_core]

In terms of user visible impact of this bug for backports:

cxl_region_invalidate_memregion() on x86 invokes wbinvd which is a
problematic instruction for virtualized environments. So, on virtualized
x86, cxl_region_invalidate_memregion() returns an error. This failure
case got missed because CXL memory-expander device passthrough is not a
production use case, and emulation of CXL devices is typically limited
to kernel development builds with CONFIG_CXL_REGION_INVALIDATION_TEST=y,
that makes cxl_region_invalidate_memregion() succeed.

In other words, the expected exposure of this bug is limited to CXL
subsystem development environments using QEMU that neglected
CONFIG_CXL_REGION_INVALIDATION_TEST=y.

Fixes: d1257d098a5a ("cxl/region: Move cache invalidation before region teardown, and before setup")
Signed-off-by: Li Zhijian <lizhijian@fujitsu.com>
Reviewed-by: Ira Weiny <ira.weiny@intel.com>
Link: https://lore.kernel.org/r/20231025085450.2514906-1-lizhijian@fujitsu.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
drivers/cxl/core/region.c