All of lore.kernel.org
 help / color / mirror / Atom feed
From: Ard Biesheuvel <ardb+git@google.com>
To: linux-kernel@vger.kernel.org
Cc: Ard Biesheuvel <ardb@kernel.org>,
	Kevin Loughlin <kevinloughlin@google.com>,
	 Tom Lendacky <thomas.lendacky@amd.com>,
	Dionna Glaze <dionnaglaze@google.com>,
	 Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
	 Dave Hansen <dave.hansen@linux.intel.com>,
	Andy Lutomirski <luto@kernel.org>,  Arnd Bergmann <arnd@arndb.de>,
	Nathan Chancellor <nathan@kernel.org>,
	 Nick Desaulniers <ndesaulniers@google.com>,
	Justin Stitt <justinstitt@google.com>,
	 Kees Cook <keescook@chromium.org>,
	Brian Gerst <brgerst@gmail.com>,
	linux-arch@vger.kernel.org,  llvm@lists.linux.dev
Subject: [PATCH v4 02/11] x86/startup_64: Replace pointer fixups with RIP-relative references
Date: Tue, 13 Feb 2024 13:41:46 +0100	[thread overview]
Message-ID: <20240213124143.1484862-15-ardb+git@google.com> (raw)
In-Reply-To: <20240213124143.1484862-13-ardb+git@google.com>

From: Ard Biesheuvel <ardb@kernel.org>

The code in __startup64() runs from a 1:1 mapping of memory, and uses
explicit pointer arithmetic to convert symbol references generated by
the compiler into references that work correctly via this 1:1 mapping.

This relies on the compiler generating absolute symbol references, which
will be resolved by the linker using the kernel virtual mapping.
However, the compiler may just as well emit RIP-relative references,
even when not operating in PIC mode, and so this explicit pointer
arithmetic is fragile and should be avoided. The fixup routines also
take a 'physical base' argument which needs to be passed around as
well.

Convert these pointer fixups to RIP-relative references, which are
guaranteed to produce the correct values without any explicit
arithmetic, removing the need to pass around the physical load address.
It also makes the code substantially easier to understand.

Replace bare 510/511 constants with the appropriate symbolic constants
while at it.  Note that pgd_index(__START_KERNEL_map) always produces
the value 511, regardless of the number of paging levels used, so a
symbolic constant is used here as well.

The remaining fixup_int()/fixup_long() calls related to 5-level paging
will be removed in a subsequent patch.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
---
 arch/x86/kernel/head64.c | 57 ++++++++------------
 1 file changed, 21 insertions(+), 36 deletions(-)

diff --git a/arch/x86/kernel/head64.c b/arch/x86/kernel/head64.c
index 9d7f12829f2d..4b08e321d168 100644
--- a/arch/x86/kernel/head64.c
+++ b/arch/x86/kernel/head64.c
@@ -77,6 +77,7 @@ static struct desc_struct startup_gdt[GDT_ENTRIES] __initdata = {
 	[GDT_ENTRY_KERNEL_DS]           = GDT_ENTRY_INIT(DESC_DATA64, 0, 0xfffff),
 };
 
+#ifdef CONFIG_X86_5LEVEL
 static void __head *fixup_pointer(void *ptr, unsigned long physaddr)
 {
 	return ptr - (void *)_text + (void *)physaddr;
@@ -87,7 +88,6 @@ static unsigned long __head *fixup_long(void *ptr, unsigned long physaddr)
 	return fixup_pointer(ptr, physaddr);
 }
 
-#ifdef CONFIG_X86_5LEVEL
 static unsigned int __head *fixup_int(void *ptr, unsigned long physaddr)
 {
 	return fixup_pointer(ptr, physaddr);
@@ -164,22 +164,21 @@ static unsigned long __head sme_postprocess_startup(struct boot_params *bp, pmdv
 /* Code in __startup_64() can be relocated during execution, but the compiler
  * doesn't have to generate PC-relative relocations when accessing globals from
  * that function. Clang actually does not generate them, which leads to
- * boot-time crashes. To work around this problem, every global pointer must
- * be adjusted using fixup_pointer().
+ * boot-time crashes. To work around this problem, every global variable must
+ * be accessed using RIP_REL_REF().
  */
 unsigned long __head __startup_64(unsigned long physaddr,
 				  struct boot_params *bp)
 {
-	unsigned long load_delta, *p;
+	pmd_t (*early_pgts)[PTRS_PER_PMD] = RIP_REL_REF(early_dynamic_pgts);
 	unsigned long pgtable_flags;
+	unsigned long load_delta;
 	pgdval_t *pgd;
 	p4dval_t *p4d;
 	pudval_t *pud;
 	pmdval_t *pmd, pmd_entry;
-	pteval_t *mask_ptr;
 	bool la57;
 	int i;
-	unsigned int *next_pgt_ptr;
 
 	la57 = check_la57_support(physaddr);
 
@@ -192,6 +191,7 @@ unsigned long __head __startup_64(unsigned long physaddr,
 	 * and the address I am actually running at.
 	 */
 	load_delta = physaddr - (unsigned long)(_text - __START_KERNEL_map);
+	RIP_REL_REF(phys_base) = load_delta;
 
 	/* Is the address not 2M aligned? */
 	if (load_delta & ~PMD_MASK)
@@ -201,25 +201,19 @@ unsigned long __head __startup_64(unsigned long physaddr,
 	load_delta += sme_get_me_mask();
 
 	/* Fixup the physical addresses in the page table */
-
-	pgd = fixup_pointer(early_top_pgt, physaddr);
-	p = pgd + pgd_index(__START_KERNEL_map);
-	if (la57)
-		*p = (unsigned long)level4_kernel_pgt;
-	else
-		*p = (unsigned long)level3_kernel_pgt;
-	*p += _PAGE_TABLE_NOENC - __START_KERNEL_map + load_delta;
-
 	if (la57) {
-		p4d = fixup_pointer(level4_kernel_pgt, physaddr);
-		p4d[511] += load_delta;
+		p4d = (p4dval_t *)&RIP_REL_REF(level4_kernel_pgt);
+		p4d[MAX_PTRS_PER_P4D - 1] += load_delta;
 	}
 
-	pud = fixup_pointer(level3_kernel_pgt, physaddr);
-	pud[510] += load_delta;
-	pud[511] += load_delta;
+	pud = &RIP_REL_REF(level3_kernel_pgt)->pud;
+	pud[PTRS_PER_PUD - 2] += load_delta;
+	pud[PTRS_PER_PUD - 1] += load_delta;
+
+	pgd = &RIP_REL_REF(early_top_pgt)->pgd;
+	pgd[PTRS_PER_PGD - 1] = (pgdval_t)(la57 ? p4d : pud) | _PAGE_TABLE_NOENC;
 
-	pmd = fixup_pointer(level2_fixmap_pgt, physaddr);
+	pmd = &RIP_REL_REF(level2_fixmap_pgt)->pmd;
 	for (i = FIXMAP_PMD_TOP; i > FIXMAP_PMD_TOP - FIXMAP_PMD_NUM; i--)
 		pmd[i] += load_delta;
 
@@ -230,16 +224,14 @@ unsigned long __head __startup_64(unsigned long physaddr,
 	 * it avoids problems around wraparound.
 	 */
 
-	next_pgt_ptr = fixup_pointer(&next_early_pgt, physaddr);
-	pud = fixup_pointer(early_dynamic_pgts[(*next_pgt_ptr)++], physaddr);
-	pmd = fixup_pointer(early_dynamic_pgts[(*next_pgt_ptr)++], physaddr);
+	pud = &early_pgts[0]->pmd;
+	pmd = &early_pgts[1]->pmd;
+	p4d = &early_pgts[2]->pmd;
+	RIP_REL_REF(next_early_pgt) = 3;
 
 	pgtable_flags = _KERNPG_TABLE_NOENC + sme_get_me_mask();
 
 	if (la57) {
-		p4d = fixup_pointer(early_dynamic_pgts[(*next_pgt_ptr)++],
-				    physaddr);
-
 		i = (physaddr >> PGDIR_SHIFT) % PTRS_PER_PGD;
 		pgd[i + 0] = (pgdval_t)p4d + pgtable_flags;
 		pgd[i + 1] = (pgdval_t)p4d + pgtable_flags;
@@ -259,8 +251,7 @@ unsigned long __head __startup_64(unsigned long physaddr,
 
 	pmd_entry = __PAGE_KERNEL_LARGE_EXEC & ~_PAGE_GLOBAL;
 	/* Filter out unsupported __PAGE_KERNEL_* bits: */
-	mask_ptr = fixup_pointer(&__supported_pte_mask, physaddr);
-	pmd_entry &= *mask_ptr;
+	pmd_entry &= RIP_REL_REF(__supported_pte_mask);
 	pmd_entry += sme_get_me_mask();
 	pmd_entry +=  physaddr;
 
@@ -286,7 +277,7 @@ unsigned long __head __startup_64(unsigned long physaddr,
 	 * error, causing the BIOS to halt the system.
 	 */
 
-	pmd = fixup_pointer(level2_kernel_pgt, physaddr);
+	pmd = &RIP_REL_REF(level2_kernel_pgt)->pmd;
 
 	/* invalidate pages before the kernel image */
 	for (i = 0; i < pmd_index((unsigned long)_text); i++)
@@ -301,12 +292,6 @@ unsigned long __head __startup_64(unsigned long physaddr,
 	for (; i < PTRS_PER_PMD; i++)
 		pmd[i] &= ~_PAGE_PRESENT;
 
-	/*
-	 * Fixup phys_base - remove the memory encryption mask to obtain
-	 * the true physical address.
-	 */
-	*fixup_long(&phys_base, physaddr) += load_delta - sme_get_me_mask();
-
 	return sme_postprocess_startup(bp, pmd);
 }
 
-- 
2.43.0.687.g38aa6559b0-goog


  parent reply	other threads:[~2024-02-13 12:42 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-02-13 12:41 [PATCH v4 00/11] x86: Confine early 1:1 mapped startup code Ard Biesheuvel
2024-02-13 12:41 ` [PATCH v4 01/11] x86/startup_64: Simplify global variable accesses in GDT/IDT programming Ard Biesheuvel
2024-02-13 20:05   ` Borislav Petkov
2024-02-13 21:53     ` Ard Biesheuvel
2024-02-14  7:28       ` Ard Biesheuvel
2024-02-15 13:52         ` Borislav Petkov
2024-02-13 12:41 ` Ard Biesheuvel [this message]
2024-02-17 12:51   ` [PATCH v4 02/11] x86/startup_64: Replace pointer fixups with RIP-relative references Borislav Petkov
2024-02-17 13:58     ` Ard Biesheuvel
2024-02-17 16:10       ` Ard Biesheuvel
2024-02-19  9:55         ` Borislav Petkov
2024-02-19 10:45           ` Ard Biesheuvel
2024-02-19 10:01       ` Borislav Petkov
2024-02-19 10:47         ` Ard Biesheuvel
2024-02-13 12:41 ` [PATCH v4 03/11] x86/startup_64: Simplify CR4 handling in startup code Ard Biesheuvel
2024-02-19 10:32   ` Borislav Petkov
2024-02-13 12:41 ` [PATCH v4 04/11] x86/startup_64: Defer assignment of 5-level paging global variables Ard Biesheuvel
2024-02-20 18:45   ` Borislav Petkov
2024-02-20 23:33     ` Ard Biesheuvel
2024-02-21 10:09       ` Borislav Petkov
2024-02-21 10:20         ` Ard Biesheuvel
2024-02-21 11:12           ` Borislav Petkov
2024-02-21 11:21             ` Ard Biesheuvel
2024-02-21 11:23               ` Borislav Petkov
2024-02-13 12:41 ` [PATCH v4 05/11] x86/startup_64: Simplify calculation of initial page table address Ard Biesheuvel
2024-02-13 12:41 ` [PATCH v4 06/11] x86/startup_64: Simplify virtual switch on primary boot Ard Biesheuvel
2024-02-13 12:41 ` [PATCH v4 07/11] efi/libstub: Add generic support for parsing mem_encrypt= Ard Biesheuvel
2024-02-19 17:00   ` Tom Lendacky
2024-02-19 17:06     ` Ard Biesheuvel
2024-02-20 19:28       ` Tom Lendacky
2024-02-13 12:41 ` [PATCH v4 08/11] x86/boot: Move mem_encrypt= parsing to the decompressor Ard Biesheuvel
2024-02-13 12:41 ` [PATCH v4 09/11] x86/sme: Move early SME kernel encryption handling into .head.text Ard Biesheuvel
2024-02-13 12:41 ` [PATCH v4 10/11] x86/sev: Move early startup code into .head.text section Ard Biesheuvel
2024-02-13 12:41 ` [PATCH v4 11/11] x86/startup_64: Drop global variables keeping track of LA57 state Ard Biesheuvel
2024-02-14 15:24   ` Brian Gerst
2024-02-14 15:44     ` Ard Biesheuvel
2024-02-14 20:25       ` Brian Gerst

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240213124143.1484862-15-ardb+git@google.com \
    --to=ardb+git@google.com \
    --cc=ardb@kernel.org \
    --cc=arnd@arndb.de \
    --cc=bp@alien8.de \
    --cc=brgerst@gmail.com \
    --cc=dave.hansen@linux.intel.com \
    --cc=dionnaglaze@google.com \
    --cc=justinstitt@google.com \
    --cc=keescook@chromium.org \
    --cc=kevinloughlin@google.com \
    --cc=linux-arch@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=llvm@lists.linux.dev \
    --cc=luto@kernel.org \
    --cc=mingo@redhat.com \
    --cc=nathan@kernel.org \
    --cc=ndesaulniers@google.com \
    --cc=tglx@linutronix.de \
    --cc=thomas.lendacky@amd.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.