| Message ID | 20190817024629.26611-4-pasha.tatashin@soleen.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | arm64: MMU enabled kexec relocation | expand |
On Fri, Aug 16, 2019 at 10:46:18PM -0400, Pavel Tatashin wrote: > trans_table_create_copy() and trans_table_map_page() are going to be > the basis for public interface of new subsystem that handles page > tables for cases which are between kernels: kexec, and hibernate. While the architecture uses the term 'translation table', in the kernel we generally use 'pgdir' or 'pgd' to refer to the tables, so please keep to that naming scheme. For example, in arch/arm64/mm/mmu.c we have a somewhat analagous function called create_pgd_mapping() -- could we use that here, to crate the mapping? Thanks, Mark. > > Signed-off-by: Pavel Tatashin <pasha.tatashin@soleen.com> > --- > arch/arm64/kernel/hibernate.c | 96 ++++++++++++++++++++++------------- > 1 file changed, 61 insertions(+), 35 deletions(-) > > diff --git a/arch/arm64/kernel/hibernate.c b/arch/arm64/kernel/hibernate.c > index 96b6f8da7e49..449d69b5651c 100644 > --- a/arch/arm64/kernel/hibernate.c > +++ b/arch/arm64/kernel/hibernate.c > @@ -182,39 +182,15 @@ int arch_hibernation_header_restore(void *addr) > } > EXPORT_SYMBOL(arch_hibernation_header_restore); > > -/* > - * Copies length bytes, starting at src_start into an new page, > - * perform cache maintentance, then maps it at the specified address low > - * address as executable. > - * > - * This is used by hibernate to copy the code it needs to execute when > - * overwriting the kernel text. This function generates a new set of page > - * tables, which it loads into ttbr0. > - * > - * Length is provided as we probably only want 4K of data, even on a 64K > - * page system. > - */ > -static int create_safe_exec_page(void *src_start, size_t length, > - unsigned long dst_addr, > - phys_addr_t *phys_dst_addr) > +int trans_table_map_page(pgd_t *trans_table, void *page, > + unsigned long dst_addr, > + pgprot_t pgprot) > { > - void *page = (void *)get_safe_page(GFP_ATOMIC); > - pgd_t *trans_table; > pgd_t *pgdp; > pud_t *pudp; > pmd_t *pmdp; > pte_t *ptep; > > - if (!page) > - return -ENOMEM; > - > - memcpy((void *)page, src_start, length); > - __flush_icache_range((unsigned long)page, (unsigned long)page + length); > - > - trans_table = (void *)get_safe_page(GFP_ATOMIC); > - if (!trans_table) > - return -ENOMEM; > - > pgdp = pgd_offset_raw(trans_table, dst_addr); > if (pgd_none(READ_ONCE(*pgdp))) { > pudp = (void *)get_safe_page(GFP_ATOMIC); > @@ -242,6 +218,44 @@ static int create_safe_exec_page(void *src_start, size_t length, > ptep = pte_offset_kernel(pmdp, dst_addr); > set_pte(ptep, pfn_pte(virt_to_pfn(page), PAGE_KERNEL_EXEC)); > > + return 0; > +} > + > +/* > + * Copies length bytes, starting at src_start into an new page, > + * perform cache maintentance, then maps it at the specified address low > + * address as executable. > + * > + * This is used by hibernate to copy the code it needs to execute when > + * overwriting the kernel text. This function generates a new set of page > + * tables, which it loads into ttbr0. > + * > + * Length is provided as we probably only want 4K of data, even on a 64K > + * page system. > + */ > +static int create_safe_exec_page(void *src_start, size_t length, > + unsigned long dst_addr, > + phys_addr_t *phys_dst_addr) > +{ > + void *page = (void *)get_safe_page(GFP_ATOMIC); > + pgd_t *trans_table; > + int rc; > + > + if (!page) > + return -ENOMEM; > + > + memcpy(page, src_start, length); > + __flush_icache_range((unsigned long)page, (unsigned long)page + length); > + > + trans_table = (void *)get_safe_page(GFP_ATOMIC); > + if (!trans_table) > + return -ENOMEM; > + > + rc = trans_table_map_page(trans_table, page, dst_addr, > + PAGE_KERNEL_EXEC); > + if (rc) > + return rc; > + > /* > * Load our new page tables. A strict BBM approach requires that we > * ensure that TLBs are free of any entries that may overlap with the > @@ -259,7 +273,7 @@ static int create_safe_exec_page(void *src_start, size_t length, > write_sysreg(phys_to_ttbr(virt_to_phys(trans_table)), ttbr0_el1); > isb(); > > - *phys_dst_addr = virt_to_phys((void *)page); > + *phys_dst_addr = virt_to_phys(page); > > return 0; > } > @@ -462,6 +476,24 @@ static int copy_page_tables(pgd_t *dst_pgdp, unsigned long start, > return 0; > } > > +int trans_table_create_copy(pgd_t **dst_pgdp, unsigned long start, > + unsigned long end) > +{ > + int rc; > + pgd_t *trans_table = (pgd_t *)get_safe_page(GFP_ATOMIC); > + > + if (!trans_table) { > + pr_err("Failed to allocate memory for temporary page tables.\n"); > + return -ENOMEM; > + } > + > + rc = copy_page_tables(trans_table, start, end); > + if (!rc) > + *dst_pgdp = trans_table; > + > + return rc; > +} > + > /* > * Setup then Resume from the hibernate image using swsusp_arch_suspend_exit(). > * > @@ -483,13 +515,7 @@ int swsusp_arch_resume(void) > * Create a second copy of just the linear map, and use this when > * restoring. > */ > - tmp_pg_dir = (pgd_t *)get_safe_page(GFP_ATOMIC); > - if (!tmp_pg_dir) { > - pr_err("Failed to allocate memory for temporary page tables.\n"); > - rc = -ENOMEM; > - goto out; > - } > - rc = copy_page_tables(tmp_pg_dir, PAGE_OFFSET, 0); > + rc = trans_table_create_copy(&tmp_pg_dir, PAGE_OFFSET, 0); > if (rc) > goto out; > > -- > 2.22.1 >
On Mon, Aug 19, 2019 at 11:58 AM Mark Rutland <mark.rutland@arm.com> wrote: > > On Fri, Aug 16, 2019 at 10:46:18PM -0400, Pavel Tatashin wrote: > > trans_table_create_copy() and trans_table_map_page() are going to be > > the basis for public interface of new subsystem that handles page > > tables for cases which are between kernels: kexec, and hibernate. > > While the architecture uses the term 'translation table', in the kernel > we generally use 'pgdir' or 'pgd' to refer to the tables, so please keep > to that naming scheme. The idea is to have a unique name space for new subsystem of page tables that are used between kernels: between stage 1 and stage 2 kexec kernel, and similarly between kernels during hibernate boot process. I picked: "trans_table" that stands for transitional page table: meaning they are used only during transition between worlds. All public functions in this subsystem will have trans_table_* prefix, and page directory will be named: "trans_table". If this is confusing, I can either use a different prefix, or describe what "trans_table" stand for in trans_table.h/.c Thank you, Pasha
On Mon, Aug 19, 2019 at 12:33:31PM -0400, Pavel Tatashin wrote: > On Mon, Aug 19, 2019 at 11:58 AM Mark Rutland <mark.rutland@arm.com> wrote: > > On Fri, Aug 16, 2019 at 10:46:18PM -0400, Pavel Tatashin wrote: > > > trans_table_create_copy() and trans_table_map_page() are going to be > > > the basis for public interface of new subsystem that handles page > > > tables for cases which are between kernels: kexec, and hibernate. > > > > While the architecture uses the term 'translation table', in the kernel > > we generally use 'pgdir' or 'pgd' to refer to the tables, so please keep > > to that naming scheme. > > The idea is to have a unique name space for new subsystem of page > tables that are used between kernels: > between stage 1 and stage 2 kexec kernel, and similarly between > kernels during hibernate boot process. > > I picked: "trans_table" that stands for transitional page table: > meaning they are used only during transition between worlds. > > All public functions in this subsystem will have trans_table_* prefix, > and page directory will be named: "trans_table". If this is confusing, > I can either use a different prefix, or describe what "trans_table" > stand for in trans_table.h/.c Ok. I think that "trans_table" is unfortunately confusing, as it clashes with the architecture terminology, and differs from what we have elsewhere. I think that "trans_pgd" would be better, as that better aligns with what we have elsewhere, and avoids the ambiguity. Thanks, Mark.
> > > While the architecture uses the term 'translation table', in the kernel > > > we generally use 'pgdir' or 'pgd' to refer to the tables, so please keep > > > to that naming scheme. > > > > The idea is to have a unique name space for new subsystem of page > > tables that are used between kernels: > > between stage 1 and stage 2 kexec kernel, and similarly between > > kernels during hibernate boot process. > > > > I picked: "trans_table" that stands for transitional page table: > > meaning they are used only during transition between worlds. > > > > All public functions in this subsystem will have trans_table_* prefix, > > and page directory will be named: "trans_table". If this is confusing, > > I can either use a different prefix, or describe what "trans_table" > > stand for in trans_table.h/.c > > Ok. > > I think that "trans_table" is unfortunately confusing, as it clashes > with the architecture terminology, and differs from what we have > elsewhere. > > I think that "trans_pgd" would be better, as that better aligns with > what we have elsewhere, and avoids the ambiguity. > Sounds good. I will rename trans_table* with trans_pgd*, and will also add a note to the comments explaining what it stands for. Thank you, Pasha
diff --git a/arch/arm64/kernel/hibernate.c b/arch/arm64/kernel/hibernate.c index 96b6f8da7e49..449d69b5651c 100644 --- a/arch/arm64/kernel/hibernate.c +++ b/arch/arm64/kernel/hibernate.c @@ -182,39 +182,15 @@ int arch_hibernation_header_restore(void *addr) } EXPORT_SYMBOL(arch_hibernation_header_restore); -/* - * Copies length bytes, starting at src_start into an new page, - * perform cache maintentance, then maps it at the specified address low - * address as executable. - * - * This is used by hibernate to copy the code it needs to execute when - * overwriting the kernel text. This function generates a new set of page - * tables, which it loads into ttbr0. - * - * Length is provided as we probably only want 4K of data, even on a 64K - * page system. - */ -static int create_safe_exec_page(void *src_start, size_t length, - unsigned long dst_addr, - phys_addr_t *phys_dst_addr) +int trans_table_map_page(pgd_t *trans_table, void *page, + unsigned long dst_addr, + pgprot_t pgprot) { - void *page = (void *)get_safe_page(GFP_ATOMIC); - pgd_t *trans_table; pgd_t *pgdp; pud_t *pudp; pmd_t *pmdp; pte_t *ptep; - if (!page) - return -ENOMEM; - - memcpy((void *)page, src_start, length); - __flush_icache_range((unsigned long)page, (unsigned long)page + length); - - trans_table = (void *)get_safe_page(GFP_ATOMIC); - if (!trans_table) - return -ENOMEM; - pgdp = pgd_offset_raw(trans_table, dst_addr); if (pgd_none(READ_ONCE(*pgdp))) { pudp = (void *)get_safe_page(GFP_ATOMIC); @@ -242,6 +218,44 @@ static int create_safe_exec_page(void *src_start, size_t length, ptep = pte_offset_kernel(pmdp, dst_addr); set_pte(ptep, pfn_pte(virt_to_pfn(page), PAGE_KERNEL_EXEC)); + return 0; +} + +/* + * Copies length bytes, starting at src_start into an new page, + * perform cache maintentance, then maps it at the specified address low + * address as executable. + * + * This is used by hibernate to copy the code it needs to execute when + * overwriting the kernel text. This function generates a new set of page + * tables, which it loads into ttbr0. + * + * Length is provided as we probably only want 4K of data, even on a 64K + * page system. + */ +static int create_safe_exec_page(void *src_start, size_t length, + unsigned long dst_addr, + phys_addr_t *phys_dst_addr) +{ + void *page = (void *)get_safe_page(GFP_ATOMIC); + pgd_t *trans_table; + int rc; + + if (!page) + return -ENOMEM; + + memcpy(page, src_start, length); + __flush_icache_range((unsigned long)page, (unsigned long)page + length); + + trans_table = (void *)get_safe_page(GFP_ATOMIC); + if (!trans_table) + return -ENOMEM; + + rc = trans_table_map_page(trans_table, page, dst_addr, + PAGE_KERNEL_EXEC); + if (rc) + return rc; + /* * Load our new page tables. A strict BBM approach requires that we * ensure that TLBs are free of any entries that may overlap with the @@ -259,7 +273,7 @@ static int create_safe_exec_page(void *src_start, size_t length, write_sysreg(phys_to_ttbr(virt_to_phys(trans_table)), ttbr0_el1); isb(); - *phys_dst_addr = virt_to_phys((void *)page); + *phys_dst_addr = virt_to_phys(page); return 0; } @@ -462,6 +476,24 @@ static int copy_page_tables(pgd_t *dst_pgdp, unsigned long start, return 0; } +int trans_table_create_copy(pgd_t **dst_pgdp, unsigned long start, + unsigned long end) +{ + int rc; + pgd_t *trans_table = (pgd_t *)get_safe_page(GFP_ATOMIC); + + if (!trans_table) { + pr_err("Failed to allocate memory for temporary page tables.\n"); + return -ENOMEM; + } + + rc = copy_page_tables(trans_table, start, end); + if (!rc) + *dst_pgdp = trans_table; + + return rc; +} + /* * Setup then Resume from the hibernate image using swsusp_arch_suspend_exit(). * @@ -483,13 +515,7 @@ int swsusp_arch_resume(void) * Create a second copy of just the linear map, and use this when * restoring. */ - tmp_pg_dir = (pgd_t *)get_safe_page(GFP_ATOMIC); - if (!tmp_pg_dir) { - pr_err("Failed to allocate memory for temporary page tables.\n"); - rc = -ENOMEM; - goto out; - } - rc = copy_page_tables(tmp_pg_dir, PAGE_OFFSET, 0); + rc = trans_table_create_copy(&tmp_pg_dir, PAGE_OFFSET, 0); if (rc) goto out;
trans_table_create_copy() and trans_table_map_page() are going to be the basis for public interface of new subsystem that handles page tables for cases which are between kernels: kexec, and hibernate. Signed-off-by: Pavel Tatashin <pasha.tatashin@soleen.com> --- arch/arm64/kernel/hibernate.c | 96 ++++++++++++++++++++++------------- 1 file changed, 61 insertions(+), 35 deletions(-)