RE: [RFC PATCH] ACPI / EC: Fix an order issue in ec_remove_handlers()
From: Zheng, Lv
Date: Thu Jul 07 2016 - 00:39:48 EST
This one and the previous one contain problems.
A patch marked as [UPDATE RFC v2] is the correct fix.
Sorry for the noise.
Thanks and best regards
-Lv
> From: Zheng, Lv
> Subject: [RFC PATCH] ACPI / EC: Fix an order issue in ec_remove_handlers()
>
> There is an order issue in ec_remove_handlers() that the functions invoked
> in it are not invoked in the reversed order of their appearance in
> ec_install_handlers(). This existing issue has been triggered by the
> following commit:
> Commit: dcf15cbded656a12335bc4151f3f75f10080a375
> Subject: ACPI / EC: Fix a boot EC regresion by restoring boot EC
> The commit invokes ec_remove_handlers() during runtime, thus uncovers
> this
> issue. This patch fixes this regression.
>
> Fixes: dcf15cbded65 ("ACPI / EC: Fix a boot EC regresion by restoring boot
> EC")
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=102421
> Reported-by: Wolfram Sang <wsa@xxxxxxxxxxxxx>
> Reported-by: Nicholas <nkudriavtsev@xxxxxxxxx>
> Cc: Wolfram Sang <wsa@xxxxxxxxxxxxx>
> Cc: Nicholas <nkudriavtsev@xxxxxxxxx>
> Signed-off-by: Lv Zheng <lv.zheng@xxxxxxxxx>
> ---
> drivers/acpi/ec.c | 14 +++++++-------
> 1 file changed, 7 insertions(+), 7 deletions(-)
>
> Index: linux-acpica/drivers/acpi/ec.c
> ==========================================================
> =========
> --- linux-acpica.orig/drivers/acpi/ec.c
> +++ linux-acpica/drivers/acpi/ec.c
> @@ -1331,8 +1331,6 @@ static int ec_install_handlers(struct ac
>
> static void ec_remove_handlers(struct acpi_ec *ec)
> {
> - acpi_ec_stop(ec, false);
> -
> if (test_bit(EC_FLAGS_EC_HANDLER_INSTALLED, &ec->flags)) {
> if (ACPI_FAILURE(acpi_remove_address_space_handler(ec-
> >handle,
> ACPI_ADR_SPACE_EC,
> &acpi_ec_space_handler)))
> @@ -1340,6 +1338,17 @@ static void ec_remove_handlers(struct ac
> clear_bit(EC_FLAGS_EC_HANDLER_INSTALLED, &ec->flags);
> }
>
> + /*
> + * Disabling EC (transactions) after removing the operation region
> + * handler. This order is required because _REG(DISCONNECT) may
> + * access the EmbeddedControl operation regions.
> + *
> + * Flushing transactions before removing the GPE handler. This is
> + * required by the current ACPICA GPE design. ACPICA GPE will
> block
> + * a GPE if there is no way to handle it.
> + */
> + acpi_ec_stop(ec, false);
> +
> if (test_bit(EC_FLAGS_GPE_HANDLER_INSTALLED, &ec->flags)) {
> if (ACPI_FAILURE(acpi_remove_gpe_handler(NULL, ec->gpe,
> &acpi_ec_gpe_handler)))