Re: [PATCH v6 08/14] x86: Secure Launch kernel late boot stub

[Jarkko Sakkinen]

On Thu May 4, 2023 at 5:50 PM EEST, Ross Philipson wrote:
> The routine slaunch_setup is called out of the x86 specific setup_arch
> routine during early kernel boot. After determining what platform is
> present, various operations specific to that platform occur. This
> includes finalizing setting for the platform late launch and verifying
> that memory protections are in place.
>
> For TXT, this code also reserves the original compressed kernel setup
> area where the APs were left looping so that this memory cannot be used.
>
> Signed-off-by: Ross Philipson <ross.philipson@xxxxxxxxxx>
> ---
>  arch/x86/kernel/Makefile   |   1 +
>  arch/x86/kernel/setup.c    |   3 +
>  arch/x86/kernel/slaunch.c  | 497 +++++++++++++++++++++++++++++++++++++++++++++
>  drivers/iommu/intel/dmar.c |   4 +
>  4 files changed, 505 insertions(+)
>  create mode 100644 arch/x86/kernel/slaunch.c
>
> diff --git a/arch/x86/kernel/Makefile b/arch/x86/kernel/Makefile
> index dd61752..3d2a33e 100644
> --- a/arch/x86/kernel/Makefile
> +++ b/arch/x86/kernel/Makefile
> @@ -72,6 +72,7 @@ obj-$(CONFIG_X86_32)		+= tls.o
>  obj-$(CONFIG_IA32_EMULATION)	+= tls.o
>  obj-y				+= step.o
>  obj-$(CONFIG_INTEL_TXT)		+= tboot.o
> +obj-$(CONFIG_SECURE_LAUNCH)	+= slaunch.o
>  obj-$(CONFIG_ISA_DMA_API)	+= i8237.o
>  obj-y				+= stacktrace.o
>  obj-y				+= cpu/
> diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
> index 16babff..592c09e 100644
> --- a/arch/x86/kernel/setup.c
> +++ b/arch/x86/kernel/setup.c
> @@ -21,6 +21,7 @@
>  #include <linux/root_dev.h>
>  #include <linux/hugetlb.h>
>  #include <linux/tboot.h>
> +#include <linux/slaunch.h>

Only because of pure curiosity: what made you choose this location in
the middle for the new include? :-)

>  #include <linux/usb/xhci-dbgp.h>
>  #include <linux/static_call.h>
>  #include <linux/swiotlb.h>
> @@ -1063,6 +1064,8 @@ void __init setup_arch(char **cmdline_p)
>  	early_gart_iommu_check();
>  #endif
>  
> +	slaunch_setup_txt();
> +
>  	/*
>  	 * partially used pages are not usable - thus
>  	 * we are rounding upwards:
> diff --git a/arch/x86/kernel/slaunch.c b/arch/x86/kernel/slaunch.c
> new file mode 100644
> index 0000000..7dba088
> --- /dev/null
> +++ b/arch/x86/kernel/slaunch.c
> @@ -0,0 +1,497 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Secure Launch late validation/setup and finalization support.
> + *
> + * Copyright (c) 2022, Oracle and/or its affiliates.
> + */
> +
> +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
> +
> +#include <linux/linkage.h>
> +#include <linux/mm.h>
> +#include <linux/io.h>
> +#include <linux/uaccess.h>
> +#include <linux/security.h>
> +#include <linux/memblock.h>
> +#include <asm/segment.h>
> +#include <asm/sections.h>
> +#include <asm/tlbflush.h>
> +#include <asm/e820/api.h>
> +#include <asm/setup.h>
> +#include <linux/slr_table.h>
> +#include <linux/slaunch.h>
> +
> +static u32 sl_flags;
> +static struct sl_ap_wake_info ap_wake_info;
> +static u64 evtlog_addr;
> +static u32 evtlog_size;
> +static u64 vtd_pmr_lo_size;
> +
> +/* This should be plenty of room */
> +static u8 txt_dmar[PAGE_SIZE] __aligned(16);
> +
> +u32 slaunch_get_flags(void)
> +{
> +	return sl_flags;
> +}
> +EXPORT_SYMBOL(slaunch_get_flags);
> +
> +struct sl_ap_wake_info *slaunch_get_ap_wake_info(void)
> +{
> +	return &ap_wake_info;
> +}
> +
> +struct acpi_table_header *slaunch_get_dmar_table(struct acpi_table_header *dmar)
> +{
> +	/* The DMAR is only stashed and provided via TXT on Intel systems */
> +	if (memcmp(txt_dmar, "DMAR", 4))
> +		return dmar;
> +
> +	return (struct acpi_table_header *)(&txt_dmar[0]);
> +}
> +
> +void __noreturn slaunch_txt_reset(void __iomem *txt,
> +				  const char *msg, u64 error)
> +{
> +	u64 one = 1, val;
> +
> +	pr_err("%s", msg);
> +
> +	/*
> +	 * This performs a TXT reset with a sticky error code. The reads of
> +	 * TXT_CR_E2STS act as barriers.
> +	 */
> +	memcpy_toio(txt + TXT_CR_ERRORCODE, &error, sizeof(error));
> +	memcpy_fromio(&val, txt + TXT_CR_E2STS, sizeof(val));
> +	memcpy_toio(txt + TXT_CR_CMD_NO_SECRETS, &one, sizeof(one));
> +	memcpy_fromio(&val, txt + TXT_CR_E2STS, sizeof(val));
> +	memcpy_toio(txt + TXT_CR_CMD_UNLOCK_MEM_CONFIG, &one, sizeof(one));
> +	memcpy_fromio(&val, txt + TXT_CR_E2STS, sizeof(val));
> +	memcpy_toio(txt + TXT_CR_CMD_RESET, &one, sizeof(one));
> +
> +	for ( ; ; )
> +		asm volatile ("hlt");
> +
> +	unreachable();
> +}
> +
> +/*
> + * The TXT heap is too big to map all at once with early_ioremap
> + * so it is done a table at a time.
> + */
> +static void __init *txt_early_get_heap_table(void __iomem *txt, u32 type,
> +					     u32 bytes)
> +{
> +	u64 base, size, offset = 0;
> +	void *heap;
> +	int i;
> +
> +	if (type > TXT_SINIT_TABLE_MAX)
> +		slaunch_txt_reset(txt,
> +			"Error invalid table type for early heap walk\n",
> +			SL_ERROR_HEAP_WALK);

Align with 'txt'.

> +
> +	memcpy_fromio(&base, txt + TXT_CR_HEAP_BASE, sizeof(base));
> +	memcpy_fromio(&size, txt + TXT_CR_HEAP_SIZE, sizeof(size));
> +
> +	/* Iterate over heap tables looking for table of "type" */
> +	for (i = 0; i < type; i++) {
> +		base += offset;
> +		heap = early_memremap(base, sizeof(u64));
> +		if (!heap)
> +			slaunch_txt_reset(txt,
> +				"Error early_memremap of heap for heap walk\n",
> +				SL_ERROR_HEAP_MAP);
> +
> +		offset = *((u64 *)heap);
> +
> +		/*
> +		 * After the first iteration, any offset of zero is invalid and
> +		 * implies the TXT heap is corrupted.
> +		 */
> +		if (!offset)
> +			slaunch_txt_reset(txt,
> +				"Error invalid 0 offset in heap walk\n",
> +				SL_ERROR_HEAP_ZERO_OFFSET);
> +
> +		early_memunmap(heap, sizeof(u64));
> +	}
> +
> +	/* Skip the size field at the head of each table */
> +	base += sizeof(u64);
> +	heap = early_memremap(base, bytes);
> +	if (!heap)
> +		slaunch_txt_reset(txt,
> +				  "Error early_memremap of heap section\n",
> +				  SL_ERROR_HEAP_MAP);
> +
> +	return heap;
> +}
> +
> +static void __init txt_early_put_heap_table(void *addr, unsigned long size)
> +{
> +	early_memunmap(addr, size);
> +}
> +
> +/*
> + * TXT uses a special set of VTd registers to protect all of memory from DMA
> + * until the IOMMU can be programmed to protect memory. There is the low
> + * memory PMR that can protect all memory up to 4G. The high memory PRM can
> + * be setup to protect all memory beyond 4Gb. Validate that these values cover
> + * what is expected.
> + */
> +static void __init slaunch_verify_pmrs(void __iomem *txt)
> +{
> +	struct txt_os_sinit_data *os_sinit_data;
> +	u32 field_offset, err = 0;
> +	const char *errmsg = "";
> +	unsigned long last_pfn;
> +
> +	field_offset = offsetof(struct txt_os_sinit_data, lcp_po_base);
> +	os_sinit_data = txt_early_get_heap_table(txt, TXT_OS_SINIT_DATA_TABLE,
> +						 field_offset);
> +
> +	/* Save a copy */
> +	vtd_pmr_lo_size = os_sinit_data->vtd_pmr_lo_size;
> +
> +	last_pfn = e820__end_of_ram_pfn();
> +
> +	/*
> +	 * First make sure the hi PMR covers all memory above 4G. In the
> +	 * unlikely case where there is < 4G on the system, the hi PMR will
> +	 * not be set.
> +	 */
> +	if (os_sinit_data->vtd_pmr_hi_base != 0x0ULL) {
> +		if (os_sinit_data->vtd_pmr_hi_base != 0x100000000ULL) {
> +			err = SL_ERROR_HI_PMR_BASE;
> +			errmsg =  "Error hi PMR base\n";
> +			goto out;
> +		}
> +
> +		if (PFN_PHYS(last_pfn) > os_sinit_data->vtd_pmr_hi_base +
> +		    os_sinit_data->vtd_pmr_hi_size) {
> +			err = SL_ERROR_HI_PMR_SIZE;
> +			errmsg = "Error hi PMR size\n";
> +			goto out;
> +		}
> +	}
> +
> +	/*
> +	 * Lo PMR base should always be 0. This was already checked in
> +	 * early stub.
> +	 */
> +
> +	/*
> +	 * Check that if the kernel was loaded below 4G, that it is protected
> +	 * by the lo PMR. Note this is the decompressed kernel. The ACM would
> +	 * have ensured the compressed kernel (the MLE image) was protected.
> +	 */
> +	if ((__pa_symbol(_end) < 0x100000000ULL) &&
> +	    (__pa_symbol(_end) > os_sinit_data->vtd_pmr_lo_size)) {
> +		err = SL_ERROR_LO_PMR_MLE;
> +		errmsg = "Error lo PMR does not cover MLE kernel\n";
> +	}
> +
> +	/*
> +	 * Other regions of interest like boot param, AP wake block, cmdline
> +	 * already checked for PMR coverage in the early stub code.
> +	 */
> +
> +out:
> +	txt_early_put_heap_table(os_sinit_data, field_offset);
> +
> +	if (err)
> +		slaunch_txt_reset(txt, errmsg, err);
> +}
> +
> +static void __init slaunch_txt_reserve_range(u64 base, u64 size)
> +{
> +	int type;
> +
> +	type = e820__get_entry_type(base, base + size - 1);
> +	if (type == E820_TYPE_RAM) {
> +		pr_info("memblock reserve base: %llx size: %llx\n", base, size);
> +		memblock_reserve(base, size);
> +	}
> +}
> +
> +/*
> + * For Intel, certain regions of memory must be marked as reserved by putting
> + * them on the memblock reserved list if they are not already e820 reserved.
> + * This includes:
> + *  - The TXT HEAP
> + *  - The ACM area
> + *  - The TXT private register bank
> + *  - The MDR list sent to the MLE by the ACM (see TXT specification)
> + *  (Normally the above are properly reserved by firmware but if it was not
> + *  done, reserve them now)
> + *  - The AP wake block
> + *  - TPM log external to the TXT heap
> + *
> + * Also if the low PMR doesn't cover all memory < 4G, any RAM regions above
> + * the low PMR must be reservered too.
> + */
> +static void __init slaunch_txt_reserve(void __iomem *txt)
> +{
> +	struct txt_sinit_memory_descriptor_record *mdr;
> +	struct txt_sinit_mle_data *sinit_mle_data;
> +	u64 base, size, heap_base, heap_size;
> +	u32 mdrnum, mdroffset, mdrslen;
> +	u32 field_offset, i;
> +	void *mdrs;
> +
> +	base = TXT_PRIV_CONFIG_REGS_BASE;
> +	size = TXT_PUB_CONFIG_REGS_BASE - TXT_PRIV_CONFIG_REGS_BASE;
> +	slaunch_txt_reserve_range(base, size);
> +
> +	memcpy_fromio(&heap_base, txt + TXT_CR_HEAP_BASE, sizeof(heap_base));
> +	memcpy_fromio(&heap_size, txt + TXT_CR_HEAP_SIZE, sizeof(heap_size));
> +	slaunch_txt_reserve_range(heap_base, heap_size);
> +
> +	memcpy_fromio(&base, txt + TXT_CR_SINIT_BASE, sizeof(base));
> +	memcpy_fromio(&size, txt + TXT_CR_SINIT_SIZE, sizeof(size));
> +	slaunch_txt_reserve_range(base, size);
> +
> +	field_offset = offsetof(struct txt_sinit_mle_data,
> +				sinit_vtd_dmar_table_size);
> +	sinit_mle_data = txt_early_get_heap_table(txt, TXT_SINIT_MLE_DATA_TABLE,
> +						  field_offset);
> +
> +	mdrnum = sinit_mle_data->num_of_sinit_mdrs;
> +	mdroffset = sinit_mle_data->sinit_mdrs_table_offset;
> +
> +	txt_early_put_heap_table(sinit_mle_data, field_offset);
> +
> +	if (!mdrnum)
> +		goto nomdr;
> +
> +	mdrslen = mdrnum * sizeof(struct txt_sinit_memory_descriptor_record);
> +
> +	mdrs = txt_early_get_heap_table(txt, TXT_SINIT_MLE_DATA_TABLE,
> +					mdroffset + mdrslen - 8);
> +
> +	mdr = mdrs + mdroffset - 8;
> +
> +	for (i = 0; i < mdrnum; i++, mdr++) {
> +		/* Spec says some entries can have length 0, ignore them */
> +		if (mdr->type > 0 && mdr->length > 0)
> +			slaunch_txt_reserve_range(mdr->address, mdr->length);
> +	}
> +
> +	txt_early_put_heap_table(mdrs, mdroffset + mdrslen - 8);
> +
> +nomdr:
> +	slaunch_txt_reserve_range(ap_wake_info.ap_wake_block,
> +				  ap_wake_info.ap_wake_block_size);
> +
> +	/*
> +	 * Earlier checks ensured that the event log was properly situated
> +	 * either inside the TXT heap or outside. This is a check to see if the
> +	 * event log needs to be reserved. If it is in the TXT heap, it is
> +	 * already reserved.
> +	 */
> +	if (evtlog_addr < heap_base || evtlog_addr > (heap_base + heap_size))
> +		slaunch_txt_reserve_range(evtlog_addr, evtlog_size);
> +
> +	for (i = 0; i < e820_table->nr_entries; i++) {
> +		base = e820_table->entries[i].addr;
> +		size = e820_table->entries[i].size;
> +		if ((base >= vtd_pmr_lo_size) && (base < 0x100000000ULL))
> +			slaunch_txt_reserve_range(base, size);
> +		else if ((base < vtd_pmr_lo_size) &&
> +			 (base + size > vtd_pmr_lo_size))
> +			slaunch_txt_reserve_range(vtd_pmr_lo_size,
> +						  base + size - vtd_pmr_lo_size);
> +	}
> +}
> +
> +/*
> + * TXT stashes a safe copy of the DMAR ACPI table to prevent tampering.
> + * It is stored in the TXT heap. Fetch it from there and make it available
> + * to the IOMMU driver.
> + */
> +static void __init slaunch_copy_dmar_table(void __iomem *txt)
> +{
> +	struct txt_sinit_mle_data *sinit_mle_data;
> +	u32 field_offset, dmar_size, dmar_offset;
> +	void *dmar;
> +
> +	memset(&txt_dmar, 0, PAGE_SIZE);
> +
> +	field_offset = offsetof(struct txt_sinit_mle_data,
> +				processor_scrtm_status);
> +	sinit_mle_data = txt_early_get_heap_table(txt, TXT_SINIT_MLE_DATA_TABLE,
> +						  field_offset);
> +
> +	dmar_size = sinit_mle_data->sinit_vtd_dmar_table_size;
> +	dmar_offset = sinit_mle_data->sinit_vtd_dmar_table_offset;
> +
> +	txt_early_put_heap_table(sinit_mle_data, field_offset);
> +
> +	if (!dmar_size || !dmar_offset)
> +		slaunch_txt_reset(txt,
> +				  "Error invalid DMAR table values\n",
> +				  SL_ERROR_HEAP_INVALID_DMAR);
> +
> +	if (unlikely(dmar_size > PAGE_SIZE))
> +		slaunch_txt_reset(txt,
> +				  "Error DMAR too big to store\n",
> +				  SL_ERROR_HEAP_DMAR_SIZE);
> +
> +
> +	dmar = txt_early_get_heap_table(txt, TXT_SINIT_MLE_DATA_TABLE,
> +					dmar_offset + dmar_size - 8);
> +	if (!dmar)
> +		slaunch_txt_reset(txt,
> +				  "Error early_ioremap of DMAR\n",
> +				  SL_ERROR_HEAP_DMAR_MAP);
> +
> +	memcpy(&txt_dmar[0], dmar + dmar_offset - 8, dmar_size);
> +
> +	txt_early_put_heap_table(dmar, dmar_offset + dmar_size - 8);
> +}
> +
> +/*
> + * The location of the safe AP wake code block is stored in the TXT heap.
> + * Fetch needed values here in the early init code for later use in SMP
> + * startup.
> + *
> + * Also get the TPM event log values are in the SLRT and have to be fetched.
> + * They will be put on the memblock reserve list later.
> + */
> +static void __init slaunch_fetch_values(void __iomem *txt)
> +{
> +	struct txt_os_mle_data *os_mle_data;
> +	struct slr_entry_log_info *log_info;
> +	struct slr_table *slrt;
> +	u8 *jmp_offset;
> +	u32 size;
> +
> +	os_mle_data = txt_early_get_heap_table(txt, TXT_OS_MLE_DATA_TABLE,
> +					       sizeof(*os_mle_data));
> +
> +	ap_wake_info.ap_wake_block = os_mle_data->ap_wake_block;
> +	ap_wake_info.ap_wake_block_size = os_mle_data->ap_wake_block_size;
> +
> +	jmp_offset = os_mle_data->mle_scratch + SL_SCRATCH_AP_JMP_OFFSET;
> +	ap_wake_info.ap_jmp_offset = *((u32 *)jmp_offset);
> +
> +	slrt = (struct slr_table *)early_memremap(os_mle_data->slrt, sizeof(*slrt));
> +	if (!slrt)
> +		slaunch_txt_reset(txt,
> +			"Error early_memremap of SLRT failed\n",
> +			SL_ERROR_SLRT_MAP);
> +
> +	size = slrt->size;
> +	early_memunmap(slrt, sizeof(*slrt));
> +
> +	slrt = (struct slr_table *)early_memremap(os_mle_data->slrt, size);
> +	if (!slrt)
> +		slaunch_txt_reset(txt,
> +			"Error early_memremap of SLRT failed\n",
> +			SL_ERROR_SLRT_MAP);
> +
> +	log_info = (struct slr_entry_log_info *)
> +		slr_next_entry_by_tag(slrt, NULL, SLR_ENTRY_LOG_INFO);
> +
> +	if (!log_info)
> +		slaunch_txt_reset(txt,
> +				  "SLRT missing logging info entry\n",
> +				  SL_ERROR_SLRT_MISSING_ENTRY);
> +
> +	evtlog_addr = log_info->addr;
> +	evtlog_size = log_info->size;
> +
> +	early_memunmap(slrt, size);
> +
> +	txt_early_put_heap_table(os_mle_data, sizeof(*os_mle_data));
> +}
> +
> +/*
> + * Intel TXT specific late stub setup and validation.
> + */
> +void __init slaunch_setup_txt(void)
> +{
> +	u64 one = TXT_REGVALUE_ONE, val;
> +	void __iomem *txt;
> +
> +	if (!boot_cpu_has(X86_FEATURE_SMX))
> +		return;
> +
> +	/*
> +	 * If booted through secure launch entry point, the loadflags
> +	 * option will be set.
> +	 */
> +	if (!(boot_params.hdr.loadflags & SLAUNCH_FLAG))
> +		return;
> +
> +	/*
> +	 * See if SENTER was done by reading the status register in the
> +	 * public space. If the public register space cannot be read, TXT may
> +	 * be disabled.
> +	 */
> +	txt = early_ioremap(TXT_PUB_CONFIG_REGS_BASE,
> +			    TXT_NR_CONFIG_PAGES * PAGE_SIZE);
> +	if (!txt)
> +		return;
> +
> +	memcpy_fromio(&val, txt + TXT_CR_STS, sizeof(val));
> +	early_iounmap(txt, TXT_NR_CONFIG_PAGES * PAGE_SIZE);
> +
> +	/* SENTER should have been done */
> +	if (!(val & TXT_SENTER_DONE_STS))
> +		panic("Error TXT.STS SENTER_DONE not set\n");
> +
> +	/* SEXIT should have been cleared */
> +	if (val & TXT_SEXIT_DONE_STS)
> +		panic("Error TXT.STS SEXIT_DONE set\n");
> +
> +	/* Now we want to use the private register space */
> +	txt = early_ioremap(TXT_PRIV_CONFIG_REGS_BASE,
> +			    TXT_NR_CONFIG_PAGES * PAGE_SIZE);
> +	if (!txt) {
> +		/* This is really bad, no where to go from here */
> +		panic("Error early_ioremap of TXT priv registers\n");
> +	}
> +
> +	/*
> +	 * Try to read the Intel VID from the TXT private registers to see if
> +	 * TXT measured launch happened properly and the private space is
> +	 * available.
> +	 */
> +	memcpy_fromio(&val, txt + TXT_CR_DIDVID, sizeof(val));
> +	if ((val & 0xffff) != 0x8086) {
> +		/*
> +		 * Can't do a proper TXT reset since it appears something is
> +		 * wrong even though SENTER happened and it should be in SMX
> +		 * mode.
> +		 */
> +		panic("Invalid TXT vendor ID, not in SMX mode\n");
> +	}
> +
> +	/* Set flags so subsequent code knows the status of the launch */
> +	sl_flags |= (SL_FLAG_ACTIVE|SL_FLAG_ARCH_TXT);
> +
> +	/*
> +	 * Reading the proper DIDVID from the private register space means we
> +	 * are in SMX mode and private registers are open for read/write.
> +	 */
> +
> +	/* On Intel, have to handle TPM localities via TXT */
> +	memcpy_toio(txt + TXT_CR_CMD_SECRETS, &one, sizeof(one));
> +	memcpy_fromio(&val, txt + TXT_CR_E2STS, sizeof(val));
> +	memcpy_toio(txt + TXT_CR_CMD_OPEN_LOCALITY1, &one, sizeof(one));
> +	memcpy_fromio(&val, txt + TXT_CR_E2STS, sizeof(val));
> +
> +	slaunch_fetch_values(txt);
> +
> +	slaunch_verify_pmrs(txt);
> +
> +	slaunch_txt_reserve(txt);
> +
> +	slaunch_copy_dmar_table(txt);
> +
> +	early_iounmap(txt, TXT_NR_CONFIG_PAGES * PAGE_SIZE);
> +
> +	pr_info("Intel TXT setup complete\n");
> +}
> diff --git a/drivers/iommu/intel/dmar.c b/drivers/iommu/intel/dmar.c
> index a3414af..5d1ac8d 100644
> --- a/drivers/iommu/intel/dmar.c
> +++ b/drivers/iommu/intel/dmar.c
> @@ -28,6 +28,7 @@
>  #include <linux/iommu.h>
>  #include <linux/numa.h>
>  #include <linux/limits.h>
> +#include <linux/slaunch.h>
>  #include <asm/irq_remapping.h>
>  
>  #include "iommu.h"
> @@ -660,6 +661,9 @@ static inline int dmar_walk_dmar_table(struct acpi_table_dmar *dmar,
>  	 */
>  	dmar_tbl = tboot_get_dmar_table(dmar_tbl);
>  
> +	/* If Secure Launch is active, it has similar logic */
> +	dmar_tbl = slaunch_get_dmar_table(dmar_tbl);
> +
>  	dmar = (struct acpi_table_dmar *)dmar_tbl;
>  	if (!dmar)
>  		return -ENODEV;
> -- 
> 1.8.3.1

Is it possible to test TXT in QEMU? Never done it so that's why I'm
asking.

BR, Jarkko