Re: [PATCH v3] acpi: Use access_width over bit_width for system memory accesses

From: Rafael J. Wysocki
Date: Tue Mar 05 2024 - 15:15:52 EST


On Fri, Mar 1, 2024 at 8:27 PM Jarred White
<jarredwhite@xxxxxxxxxxxxxxxxxxx> wrote:
>
> To align with ACPI 6.3+, since bit_width can be any 8-bit value, we cannot
> depend on it being always on a clean 8b boundary. This was uncovered on the
> Cobalt 100 platform.
>
> SError Interrupt on CPU26, code 0xbe000011 -- SError
> CPU: 26 PID: 1510 Comm: systemd-udevd Not tainted 5.15.2.1-13 #1
> Hardware name: MICROSOFT CORPORATION, BIOS MICROSOFT CORPORATION
> pstate: 62400009 (nZCv daif +PAN -UAO +TCO -DIT -SSBS BTYPE=--)
> pc : cppc_get_perf_caps+0xec/0x410
> lr : cppc_get_perf_caps+0xe8/0x410
> sp : ffff8000155ab730
> x29: ffff8000155ab730 x28: ffff0080139d0038 x27: ffff0080139d0078
> x26: 0000000000000000 x25: ffff0080139d0058 x24: 00000000ffffffff
> x23: ffff0080139d0298 x22: ffff0080139d0278 x21: 0000000000000000
> x20: ffff00802b251910 x19: ffff0080139d0000 x18: ffffffffffffffff
> x17: 0000000000000000 x16: ffffdc7e111bad04 x15: ffff00802b251008
> x14: ffffffffffffffff x13: ffff013f1fd63300 x12: 0000000000000006
> x11: ffffdc7e128f4420 x10: 0000000000000000 x9 : ffffdc7e111badec
> x8 : ffff00802b251980 x7 : 0000000000000000 x6 : ffff0080139d0028
> x5 : 0000000000000000 x4 : ffff0080139d0018 x3 : 00000000ffffffff
> x2 : 0000000000000008 x1 : ffff8000155ab7a0 x0 : 0000000000000000
> Kernel panic - not syncing: Asynchronous SError Interrupt
> CPU: 26 PID: 1510 Comm: systemd-udevd Not tainted
> 5.15.2.1-13 #1
> Hardware name: MICROSOFT CORPORATION, BIOS MICROSOFT CORPORATION
> Call trace:
> dump_backtrace+0x0/0x1e0
> show_stack+0x24/0x30
> dump_stack_lvl+0x8c/0xb8
> dump_stack+0x18/0x34
> panic+0x16c/0x384
> add_taint+0x0/0xc0
> arm64_serror_panic+0x7c/0x90
> arm64_is_fatal_ras_serror+0x34/0xa4
> do_serror+0x50/0x6c
> el1h_64_error_handler+0x40/0x74
> el1h_64_error+0x7c/0x80
> cppc_get_perf_caps+0xec/0x410
> cppc_cpufreq_cpu_init+0x74/0x400 [cppc_cpufreq]
> cpufreq_online+0x2dc/0xa30
> cpufreq_add_dev+0xc0/0xd4
> subsys_interface_register+0x134/0x14c
> cpufreq_register_driver+0x1b0/0x354
> cppc_cpufreq_init+0x1a8/0x1000 [cppc_cpufreq]
> do_one_initcall+0x50/0x250
> do_init_module+0x60/0x27c
> load_module+0x2300/0x2570
> __do_sys_finit_module+0xa8/0x114
> __arm64_sys_finit_module+0x2c/0x3c
> invoke_syscall+0x78/0x100
> el0_svc_common.constprop.0+0x180/0x1a0
> do_el0_svc+0x84/0xa0
> el0_svc+0x2c/0xc0
> el0t_64_sync_handler+0xa4/0x12c
> el0t_64_sync+0x1a4/0x1a8
>
> Instead, use access_width to determine the size and use the offset and width
> to shift and mask the bits we want to read/write out. Make sure to add a check
> for system memory since pcc redefines the access_width to subspace id. If
> access_width is not set, then fallback to using the bit_width.
>
> Signed-off-by: Jarred White <jarredwhite@xxxxxxxxxxxxxxxxxxx>
> CC: srivatsabhat@xxxxxxxxxxxxxxxxxxx
> CC: stable@xxxxxxxxxxxxxxx #5.15+
> ---
> changelog:
> v1-->v2:
> https://lore.kernel.org/linux-acpi/20231216001312.1160-1-jarredwhite@xxxxxxxxxxxxxxxxxxx/
> 1. Fixed coding style errors
> 2. Backwards compatibility with ioremapping of address still an
> open question. Suggestions are welcomed.
> v2-->v3:
> 1. Created a fallback mechanism to revert to using the bit_width
> 2. Re-labeled macro to GET_BIT_WIDTH
> 3. Collapsed the if/else statements in the cpc_read() and cpc_write() routines
>
> drivers/acpi/cppc_acpi.c | 32 +++++++++++++++++++++++++++-----
> 1 file changed, 27 insertions(+), 5 deletions(-)
>
> diff --git a/drivers/acpi/cppc_acpi.c b/drivers/acpi/cppc_acpi.c
> index d155a86a8614..57de7edda7a5 100644
> --- a/drivers/acpi/cppc_acpi.c
> +++ b/drivers/acpi/cppc_acpi.c
> @@ -166,6 +166,13 @@ show_cppc_data(cppc_get_perf_caps, cppc_perf_caps, nominal_freq);
> show_cppc_data(cppc_get_perf_ctrs, cppc_perf_fb_ctrs, reference_perf);
> show_cppc_data(cppc_get_perf_ctrs, cppc_perf_fb_ctrs, wraparound_time);
>
> +/* Check for valid access_width, otherwise, fallback to using the bit_width */
> +#define GET_BIT_WIDTH(reg) ((reg)->access_width ? (8 << ((reg)->access_width - 1)) : (reg)->bit_width)
> +
> +/* Shift and apply the mask for CPC reads/writes */
> +#define MASK_VAL(reg, val) ((val) >> ((reg)->bit_offset & \
> + GENMASK(((reg)->bit_width), 0)))
> +
> static ssize_t show_feedback_ctrs(struct kobject *kobj,
> struct kobj_attribute *attr, char *buf)
> {
> @@ -780,6 +787,7 @@ int acpi_cppc_processor_probe(struct acpi_processor *pr)
> } else if (gas_t->space_id == ACPI_ADR_SPACE_SYSTEM_MEMORY) {
> if (gas_t->address) {
> void __iomem *addr;
> + size_t access_width;
>
> if (!osc_cpc_flexible_adr_space_confirmed) {
> pr_debug("Flexible address space capability not supported\n");
> @@ -787,7 +795,9 @@ int acpi_cppc_processor_probe(struct acpi_processor *pr)
> goto out_free;
> }
>
> - addr = ioremap(gas_t->address, gas_t->bit_width/8);
> + /* Check for access_width for the size, otherwise use bit_width */
> + access_width = GET_BIT_WIDTH(gas_t) / 8;
> + addr = ioremap(gas_t->address, access_width);
> if (!addr)
> goto out_free;
> cpc_ptr->cpc_regs[i-2].sys_mem_vaddr = addr;
> @@ -983,6 +993,7 @@ int __weak cpc_write_ffh(int cpunum, struct cpc_reg *reg, u64 val)
> static int cpc_read(int cpu, struct cpc_register_resource *reg_res, u64 *val)
> {
> void __iomem *vaddr = NULL;
> + int size;
> int pcc_ss_id = per_cpu(cpu_pcc_subspace_idx, cpu);
> struct cpc_reg *reg = &reg_res->cpc_entry.reg;
>
> @@ -994,7 +1005,7 @@ static int cpc_read(int cpu, struct cpc_register_resource *reg_res, u64 *val)
> *val = 0;
>
> if (reg->space_id == ACPI_ADR_SPACE_SYSTEM_IO) {
> - u32 width = 8 << (reg->access_width - 1);
> + u32 width = GET_BIT_WIDTH(reg);
> u32 val_u32;
> acpi_status status;
>
> @@ -1018,7 +1029,9 @@ static int cpc_read(int cpu, struct cpc_register_resource *reg_res, u64 *val)
> return acpi_os_read_memory((acpi_physical_address)reg->address,
> val, reg->bit_width);
>
> - switch (reg->bit_width) {
> + size = GET_BIT_WIDTH(reg);
> +
> + switch (size) {
> case 8:
> *val = readb_relaxed(vaddr);
> break;
> @@ -1037,18 +1050,22 @@ static int cpc_read(int cpu, struct cpc_register_resource *reg_res, u64 *val)
> return -EFAULT;
> }
>
> + if (reg->space_id == ACPI_ADR_SPACE_SYSTEM_MEMORY)
> + *val = MASK_VAL(reg, *val);
> +
> return 0;
> }
>
> static int cpc_write(int cpu, struct cpc_register_resource *reg_res, u64 val)
> {
> int ret_val = 0;
> + int size;
> void __iomem *vaddr = NULL;
> int pcc_ss_id = per_cpu(cpu_pcc_subspace_idx, cpu);
> struct cpc_reg *reg = &reg_res->cpc_entry.reg;
>
> if (reg->space_id == ACPI_ADR_SPACE_SYSTEM_IO) {
> - u32 width = 8 << (reg->access_width - 1);
> + u32 width = GET_BIT_WIDTH(reg);
> acpi_status status;
>
> status = acpi_os_write_port((acpi_io_address)reg->address,
> @@ -1070,7 +1087,12 @@ static int cpc_write(int cpu, struct cpc_register_resource *reg_res, u64 val)
> return acpi_os_write_memory((acpi_physical_address)reg->address,
> val, reg->bit_width);
>
> - switch (reg->bit_width) {
> + size = GET_BIT_WIDTH(reg);
> +
> + if (reg->space_id == ACPI_ADR_SPACE_SYSTEM_MEMORY)
> + val = MASK_VAL(reg, val);
> +
> + switch (size) {
> case 8:
> writeb_relaxed(val, vaddr);
> break;
> --

Applied as 6.9 material with some edits in the subject and changelog
and some adjustments of the new comments (one edited and one dropped).

Thanks!