Xen+hafnium fixes #7246
Xen+hafnium fixes #7246
Conversation
core/arch/arm/kernel/thread_a64.S
Outdated
| @@ -1153,6 +1153,15 @@ END_FUNC el0_sync_abort | |||
| write_apiakeylo x2 | |||
| isb | |||
| #endif | |||
|
|
|||
| #ifdef CFG_CORE_FFA | |||
| /* x0 is still pointing to the current thread_ctx | |||
There was a problem hiding this comment.
Thanks, I'll add it.
| adr_l x2, threads | ||
| madd x2, x1, x0, x2 | ||
| ldr w1, [x2, #THREAD_CTX_TSD_RPC_TARGET_INFO] | ||
| /* w1 holds rpc_target_info */ |
There was a problem hiding this comment.
maybe remove this inline comment. Implicit from above inline comment header line.
| @@ -139,6 +137,8 @@ struct initcall { | |||
| #define nex_release_init_resource(fn) release_init_resource(fn) | |||
| #endif | |||
|
|
|||
| #define boot_final(fn) __define_initcall(final, 8, fn) | |||
There was a problem hiding this comment.
Maybe swap order in inline comment at line 83:
s/boot_final() / nex_*init*()/nex_*ini / boot_final()/
I think there should be oter some updates in the inline description because some sentences are wrong or not accurate, e.g.:
"preinit_() are always called before functions registered with _init()."
"boot_final() functions are called first before exiting to normal world the first"
(edited) can be addressed later in a dedicated P-R
There was a problem hiding this comment.
I'll swap boot_final() / nex_init() in the comment above.
I'll propose an update in a later PR for the description.
|
Update |
There was a problem hiding this comment.
Reviewed-by: Etienne Carriere <[email protected]> for the series.
The 2 commit messages that mention SPMC at S-EL2 should maybe also mention SPMC at EL3 (which is possible with CFG_CORE_SEL1_SPMC=n).
Call virt_add_guest_spec_data() for the struct notif_vm_bitmap to make it accessible from notif_send_async(). Signed-off-by: Jens Wiklander <[email protected]> Reviewed-by: Etienne Carriere <[email protected]>
Make the boot_final() call last among the final call, that is, after the xen_*_init*() calls. spmc_init() accesses the manifest_fd so it must be called before release_manifest_dt() removes it. Signed-off-by: Jens Wiklander <[email protected]> Reviewed-by: Etienne Carriere <[email protected]>
Add __nex_bss and __nex_data annotations for the rxtx buffers used in a configuration with SPMC at S-EL2 or EL3. Signed-off-by: Jens Wiklander <[email protected]> Reviewed-by: Etienne Carriere <[email protected]>
thread_foreign_intr_exit() is called after the thread state has been saved and the thread is suspended. With virtualization enabled (CFG_NS_VIRTUALIZATION=y) the virt_unset_guest() is also called. After this, the guests thread contexts aren't available any longer. For FF-A thread_foreign_intr_exit() needs a few fields from the suspend threads context so extract those before suspending the thread and pass them as parameters for thread_foreign_intr_exit(). Signed-off-by: Jens Wiklander <[email protected]> Reviewed-by: Etienne Carriere <[email protected]>
jenswi-linaro
force-pushed
the
xen_haf
branch
from
d6c9740 to
74e9674
Compare
|
Comments addressed, squashed, and tags applied. |
No description provided.