On 11 March 2015 at 18:38, Andrew Jones <[email protected]> wrote: > This series fixes and extends the determination of whether or > not an address is executable for LPAE translations. The main > patch is 3/3, and describes the details in its commit message. > Patch 1/3 prepares for patch 2/3, which is prep for 3/3, and also > fixes a potential problem with checking access permissions on > v6 processors when they use the simple AP format. > > Changes since v3: > * Remove a useless is_aa64/case1/is_user code setting wxn=0 [Peter] > > Changes since v2: > * Collect NSTable bits into ns > * Explicitly check {prot,user}_rw against PAGE_READ. !{prot,user}_rw > means no read or write, but, since we can't have write without read, > it also means !({prot,user}_rw & PAGE_READ). Making this explicit > can reduce confusion when trying to determine with prot flag we > actually care about. [Peter Maydell] > > Changes since v1: > * There are now 3 patches instead of 5. This is due to approaching > the prep patches in a different way. Most notably, simple AP > format is now handled with its own function. [Peter Maydell] > * A check of SCTLR_AFE is guarded with ARM_FEATURE_V6K [Peter Maydell] > * All other cleanups suggested by Peter Maydell. > > Tested by booting Linux on mach-virt with cortex-a15 and cortex-a57 > (just up to 'Unable to mount root fs'), and also with a kvm-unit-tests > test. The curious can check out the unit test here [1]. > > Thanks in advance for reviews! > > drew > > [1] > https://github.com/rhdrjones/kvm-unit-tests/commit/ee553e4bb795b0150e31c794bf8953dfb08d619a > > Andrew Jones (3): > target-arm: convert check_ap to ap_to_rw_prot > target-arm: fix get_phys_addr_v6/SCTLR_AFE access check > target-arm: get_phys_addr_lpae: more xn control
Thanks; applied all to target-arm.next. Since this is a bugfix I'm aiming it to go into 2.3. -- PMM
