Linux To Begin Tightening Up Ability To Write To CPU MSRs From User-Space
Via the Linux kernel x86 MSR driver, writes to arbitrary model specific registers from user-space is allowed -- assuming you have root permissions. But even with requiring root access, there are security implications and other issues in allowing any CPU MSRs to be written to from user-space without the intervention of the kernel via /dev/cpu/[CPU-number]/msr.
SUSE's Borislav Petkov has added support to the x86 MSR driver to begin filtering of MSR writes. Petkov reinforced with the commit, "Having unfettered access to all MSRs on a system is and has always been a disaster waiting to happen. Think performance counter MSRs, MSRs with sticky or locked bits, MSRs making major system changes like loading microcode, MTRRs, PAT configuration, TSC counter, security mitigations MSRs, you name it. This also destroys all the kernel's caching of MSR values for performance, as the recent case with MSR_AMD64_LS_CFG showed. Another example is writing MSRs by mistake by simply typing the wrong MSR address. System freezes have been experienced that way. In general, poking at MSRs under the kernel's feet is a bad bad idea."
The new default behavior starting with Linux 5.9 will be to log messages to dmesg whenever a MSR write from user-space happens with a "write to unrecognized MSR" notification.
But this MSR filtering infrastructure also adds the ability to block MSR writes from user-space. Ultimately they will likely transition to that as the default at a later time to ensure MSR writes from user-space do not happen. This behavior right now can be toggled via the msr.allow_writes= kernel module paramrter with on/off/default. Should legitimate use-cases come up where writes to MSRs from user-space are still desired, they may add the infrastructure to selectively grant/deny access to specific MSRs and ensure they are sanitized by the kernel.
17 Comments
