On 11/27/13 at 12:44pm, Borislav Petkov wrote:
> On Tue, Nov 26, 2013 at 01:57:51PM +0800, Dave Young wrote:
> > kexec kernel will need exactly same mapping for
> > efi runtime memory ranges. Thus here export the
> > runtime ranges mapping to sysfs, kexec-tools
> > will assemble them and pass to 2nd kernel via
> > setup_data.
> >
> > Introducing a new directly /sys/firmware/efi/runtime-map
> > Just like /sys/firmware/memmap. Containing below attribute
> > in each file of that directory:
> > attribute num_pages phys_addr type virt_addr
> >
> > It will not work for efi 32bit. Only x86_64 currently.
> >
> > Matt: s/efi-runtime-map.c/runtime-map.c
> > change dir name to runtime-map
> > update to use desc_size in efi_runtime_map
> > cleaup the code, add function efi_save_runtime_map
> > improve err handling
> >
> > Signed-off-by: Dave Young <dyoung at redhat.com>
> > ---
> > .../ABI/testing/sysfs-firmware-efi-runtime-map | 45 +++++
> > arch/x86/platform/efi/efi.c | 26 +++
> > drivers/firmware/efi/Kconfig | 10 ++
> > drivers/firmware/efi/Makefile | 1 +
> > drivers/firmware/efi/efi.c | 3 +-
> > drivers/firmware/efi/runtime-map.c | 199 +++++++++++++++++++++
> > include/linux/efi.h | 6 +
> > 7 files changed, 289 insertions(+), 1 deletion(-)
> > create mode 100644 Documentation/ABI/testing/sysfs-firmware-efi-runtime-map
> > create mode 100644 drivers/firmware/efi/runtime-map.c
> >
> > diff --git a/Documentation/ABI/testing/sysfs-firmware-efi-runtime-map b/Documentation/ABI/testing/sysfs-firmware-efi-runtime-map
> > new file mode 100644
> > index 0000000..dab8d41
> > --- /dev/null
> > +++ b/Documentation/ABI/testing/sysfs-firmware-efi-runtime-map
> > @@ -0,0 +1,45 @@
> > +What: /sys/firmware/efi/runtime-map/
> > +Date: Oct 2013
> > +Contact: Dave Young <dyoung at redhat.com>
> > +Description:
> > + Switching efi runtime services to virtual mode requires
> > + that all efi memory ranges which has the runtime attribute
>
> s/has/have/
Will fix
>
> > + bit set to be mapped to virtual addresses.
> > +
> > + In kexec kernel kernel can not entering virtual mode again
> > + because there's a limitation that SetVirtualAddressMap can
> > + only be called once for entering virtual mode. But kexec
> > + kernel still need maintain same physical address to virtual
> > + address mapping as the 1st kernel.
>
> Simply say
>
> "SetVirtualAddressMap has the limitation that it can be called only once
> per system lifetime so a second kernel, like kexec, needs to reuse the
> exact same virtual addresses like the first kernel."
>
Matt are suggesting to below one, is it ok to you? I'm fine with both.
The efi runtime services can only be switched to virtual
mode once without rebooting. The kexec kernel must maintain
the same physical to virtual address mappings as the first
kernel. The mappings are exported to sysfs so userspace tools
can reassemble them and pass them into the kexec kernel.
>
> > The mappings are exported
> > + to sysfs so userspace tools can reassemble them and pass them
> > + into kexec kernel.
> > +
> > + /sys/firmware/efi/runtim-map/ is what kernel export for
> > + this purpose. The structure is as follows:
>
> "/sys/firmware/efi/runtime-map/ is the directory the kernel exports that
> information in."
Ok.
>
> > +
> > + subdirectories are named with the number of the memory range:
> > +
> > + /sys/firmware/efi/runtime-map/0
> > + /sys/firmware/efi/runtime-map/1
> > + /sys/firmware/efi/runtime-map/2
> > + /sys/firmware/efi/runtime-map/3
> > + ...
> > +
> > + Each subdirectory contains five files:
> > +
> > + attribute : The attribute of the memory range.
>
> attributes
Ok.
>
> > + num_pages : The size of the memory range in page number.
>
> in 4K pages.
Is it always 4K, how about "in pages"?
>
> > + phys_addr : The start physical address of the memory range.
>
> The physical address of the memory range
Ok.
>
> > + type : The type of the memory range.
> > + virt_addr : The start virtual address of the memory range.
>
> The virtual address...
Ok.
>
> > +
> > + Above values are all hexadecimal number with the '0x' prefix.
>
> numbers
Ok.
>
> > +
> > + So, for example:
> > +
> > + /sys/firmware/efi/runtime-map/0/attribute
> > + /sys/firmware/efi/runtime-map/0/num_pages
> > + /sys/firmware/efi/runtime-map/0/phys_addr
> > + /sys/firmware/efi/runtime-map/0/type
> > + /sys/firmware/efi/runtime-map/0/virt_addr
> > + ...
>
> What is this example supposed to show? The above is already detailed
> enough.
Ok, I can remove them.
>
> > diff --git a/arch/x86/platform/efi/efi.c b/arch/x86/platform/efi/efi.c
> > index 2a292aa..c3a2aaa 100644
> > --- a/arch/x86/platform/efi/efi.c
> > +++ b/arch/x86/platform/efi/efi.c
> > @@ -76,6 +76,9 @@ static __initdata efi_config_table_type_t arch_tables[] = {
> > {NULL_GUID, NULL, NULL},
> > };
> >
> > +void *efi_runtime_map;
> > +int nr_efi_runtime_map;
> > +
> > /*
> > * Returns 1 if 'facility' is enabled, 0 otherwise.
> > */
> > @@ -811,6 +814,21 @@ static void __init efi_merge_regions(void)
> > }
> > }
> >
> > +static int __init efi_save_runtime_map(efi_memory_desc_t *md)
>
> It is a static function so no need for the "efi_" prefix.
>
> Also, it should be called save_runtime_mem_desc or so because this is
> what it does.
Ok.
>
> > +{
> > + efi_runtime_map = krealloc(efi_runtime_map,
> > + (nr_efi_runtime_map + 1) *
> > + memmap.desc_size, GFP_KERNEL);
>
> WARNING: Reusing the krealloc arg is almost always a bug
> #21: FILE: arch/x86/platform/efi/efi.c:819:
> + efi_runtime_map = krealloc(efi_runtime_map,
>
> Please run your diffs through checkpatch.pl --strict before submitting.
>
> Btw, we use krealloc below again which should be changed too.
Since we do not know the exact number of ranges, introducing a list
or count the totoal range numbers before will add more code. Do you
really think it's necessary?
I tend to not fix every warnings from checkpatch if it's not a must-fix
one.
>
> > + if (!efi_runtime_map)
> > + return -ENOMEM;
> > +
> > + memcpy(efi_runtime_map + nr_efi_runtime_map * memmap.desc_size,
> > + md, memmap.desc_size);
> > + nr_efi_runtime_map++;
> > +
> > + return 0;
> > +}
> > +
> > /*
> > * Map efi memory ranges for runtime serivce and update new_memmap with virtual
> > * addresses.
> > @@ -821,6 +839,7 @@ static void * __init efi_map_regions(int *count)
> > void *p, *new_memmap = NULL;
> > unsigned long size;
> > u64 end, systab;
> > + int error = 0;
> >
> > for (p = memmap.map; p < memmap.map_end; p += memmap.desc_size) {
> > md = p;
> > @@ -852,9 +871,16 @@ static void * __init efi_map_regions(int *count)
> >
> > memcpy(new_memmap + (*count * memmap.desc_size), md,
> > memmap.desc_size);
> > + if (!error && md->type != EFI_BOOT_SERVICES_CODE &&
> > + md->type != EFI_BOOT_SERVICES_DATA)
> > + error = efi_save_runtime_map(md);
>
> I'm not sure I like this entangled stuff with the error code - I can
> barely understand what's going on there - not necessarily because of
> your changes but because this code is not that trivial, especially after
> time has passed.
suppose kexec runtime map preparing fails we still can get 1st kernel boot ok.
>
> And I don't see you kfreeing efi_runtime_map in the case where the first
> krealloc() fails at some iteration.
Good catch, the original krealloc error handling also need improve.
Ok, so I will improve the code by adding a link list for new_memmap
and removing the krealloc.
>
> What would be much cleaner IMO would be if you let efi_map_regions()
> finish, *and*, *only* if it was successful iterate again over the
> regions and do efi_save_runtime_map() on the runtime ones. This will
> make the code much more readable too.
For simplifying the code, will do.
>
> > (*count)++;
> > }
> >
> > + if (error) {
> > + nr_efi_runtime_map = 0;
> > + kfree(efi_runtime_map);
> > + }
>
> This error gets overwritten each time the loop runs and you possibly
> lose !0 values. What I'm suggesting above would take care of that case
> too.
>
> > ret:
> > return new_memmap;
> > }
> > diff --git a/drivers/firmware/efi/Kconfig b/drivers/firmware/efi/Kconfig
> > index b0fc7c7..b4d01ad 100644
> > --- a/drivers/firmware/efi/Kconfig
> > +++ b/drivers/firmware/efi/Kconfig
> > @@ -36,4 +36,14 @@ config EFI_VARS_PSTORE_DEFAULT_DISABLE
> > backend for pstore by default. This setting can be overridden
> > using the efivars module's pstore_disable parameter.
> >
> > +config EFI_RUNTIME_MAP
> > + bool "Export efi runtime maps to sysfs" if EXPERT
> > + default X86 && EFI
>
> This should depend on KEXEC.
I did not add KEXEC because I thought it can be used for debugging purpose.
>
> > + help
> > + Export efi runtime memory maps to /sys/firmware/efi/runtime-map.
> > + That memory map is used for example by kexec to set up efi virtual
> > + mapping the 2nd kernel, but can also be used for debugging purposes.
> > +
> > + See also Documentation/ABI/testing/sysfs-firmware-efi-runtime-map.
> > +
> > endmenu
> > diff --git a/drivers/firmware/efi/Makefile b/drivers/firmware/efi/Makefile
> > index 99245ab..ecadcc1 100644
> > --- a/drivers/firmware/efi/Makefile
> > +++ b/drivers/firmware/efi/Makefile
> > @@ -4,3 +4,4 @@
> > obj-y += efi.o vars.o
> > obj-$(CONFIG_EFI_VARS) += efivars.o
> > obj-$(CONFIG_EFI_VARS_PSTORE) += efi-pstore.o
> > +obj-$(CONFIG_EFI_RUNTIME_MAP) += runtime-map.o
>
> This will conflict with upstream:
Will check and update, what tree are you using?
>
> checking file drivers/firmware/efi/Makefile
> Hunk #1 FAILED at 4.
> 1 out of 1 hunk FAILED
>
> It should be:
>
> diff --git a/drivers/firmware/efi/Makefile b/drivers/firmware/efi/Makefile
> index 9ba156d3c775..7f05c368b868 100644
> --- a/drivers/firmware/efi/Makefile
> +++ b/drivers/firmware/efi/Makefile
> @@ -5,3 +5,4 @@ obj-y += efi.o vars.o
> obj-$(CONFIG_EFI_VARS) += efivars.o
> obj-$(CONFIG_EFI_VARS_PSTORE) += efi-pstore.o
> obj-$(CONFIG_UEFI_CPER) += cper.o
> +obj-$(CONFIG_EFI_RUNTIME_MAP) += runtime-map.o
>
>
> > diff --git a/drivers/firmware/efi/efi.c b/drivers/firmware/efi/efi.c
> > index 5d85956..a480de8 100644
> > --- a/drivers/firmware/efi/efi.c
> > +++ b/drivers/firmware/efi/efi.c
> > @@ -38,7 +38,8 @@ struct efi __read_mostly efi = {
> > };
> > EXPORT_SYMBOL(efi);
> >
> > -static struct kobject *efi_kobj;
> > +struct kobject *efi_kobj;
> > +EXPORT_SYMBOL_GPL(efi_kobj);
>
> Can we move this private to linux/drivers/efi/runtime-map.c? If yes,
> then there's no need to export it.
Hmm, it's created in efi.c so I think we can not move it..
>
> > static struct kobject *efivars_kobj;
> >
> > /*
> > diff --git a/drivers/firmware/efi/runtime-map.c b/drivers/firmware/efi/runtime-map.c
> > new file mode 100644
> > index 0000000..036d099
> > --- /dev/null
> > +++ b/drivers/firmware/efi/runtime-map.c
> > @@ -0,0 +1,199 @@
> > +/*
> > + * linux/drivers/efi/runtime-map.c
> > + * Copyright (C) 2013 Red Hat, Inc., Dave Young <dyoung at redhat.com>
> > + *
> > + * This program is free software; you can redistribute it and/or modify
> > + * it under the terms of the GNU General Public License v2.0 as published by
> > + * the Free Software Foundation
> > + *
> > + * This program is distributed in the hope that it will be useful,
> > + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
> > + * GNU General Public License for more details.
>
> Please drop this boilerplate text and do a single sentence saying that
> it is licensed under GPL v2 and refer to COPYING in the kernel repo.
Ok.
>
> > + *
> > + */
> > +
> > +#include <linux/string.h>
> > +#include <linux/kernel.h>
> > +#include <linux/module.h>
> > +#include <linux/types.h>
> > +#include <linux/efi.h>
> > +#include <linux/slab.h>
> > +
> > +#include <asm/setup.h>
> > +
> > +struct efi_runtime_map_entry {
> > + efi_memory_desc_t md;
> > + struct kobject kobj; /* kobject for each entry */
> > +};
> > +
> > +static struct efi_runtime_map_entry **map_entries;
> > +
> > +static ssize_t map_attr_show(struct kobject *kobj,
> > + struct attribute *attr, char *buf);
> > +static ssize_t type_show(struct efi_runtime_map_entry *entry, char *buf);
> > +static ssize_t phys_addr_show(struct efi_runtime_map_entry *entry, char *buf);
> > +static ssize_t virt_addr_show(struct efi_runtime_map_entry *entry, char *buf);
> > +static ssize_t num_pages_show(struct efi_runtime_map_entry *entry, char *buf);
> > +static ssize_t attribute_show(struct efi_runtime_map_entry *entry, char *buf);
> > +
> > +struct map_attribute {
> > + struct attribute attr;
> > + ssize_t (*show)(struct efi_runtime_map_entry *entry, char *buf);
> > +};
> > +
> > +static struct map_attribute map_type_attr = __ATTR_RO(type);
> > +static struct map_attribute map_phys_addr_attr = __ATTR_RO(phys_addr);
> > +static struct map_attribute map_virt_addr_attr = __ATTR_RO(virt_addr);
> > +static struct map_attribute map_num_pages_attr = __ATTR_RO(num_pages);
> > +static struct map_attribute map_attribute_attr = __ATTR_RO(attribute);
> > +
> > +/*
> > + * These are default attributes that are added for every memmap entry.
> > + */
> > +static struct attribute *def_attrs[] = {
> > + &map_type_attr.attr,
> > + &map_phys_addr_attr.attr,
> > + &map_virt_addr_attr.attr,
> > + &map_num_pages_attr.attr,
> > + &map_attribute_attr.attr,
> > + NULL
> > +};
> > +
> > +static const struct sysfs_ops map_attr_ops = {
> > + .show = map_attr_show,
> > +};
> > +
> > +static inline struct efi_runtime_map_entry *to_map_entry(struct kobject *kobj)
> > +{
> > + return container_of(kobj, struct efi_runtime_map_entry, kobj);
> > +}
> > +
> > +static void map_release(struct kobject *kobj)
> > +{
> > + struct efi_runtime_map_entry *entry;
> > +
> > + entry = to_map_entry(kobj);
> > + kfree(entry);
> > +}
> > +
> > +static struct kobj_type __refdata map_ktype = {
> > + .sysfs_ops = &map_attr_ops,
> > + .default_attrs = def_attrs,
> > + .release = map_release,
> > +};
> > +
> > +static ssize_t type_show(struct efi_runtime_map_entry *entry, char *buf)
> > +{
> > + return snprintf(buf, PAGE_SIZE, "0x%x\n", entry->md.type);
> > +}
> > +
> > +static ssize_t phys_addr_show(struct efi_runtime_map_entry *entry, char *buf)
> > +{
> > + return snprintf(buf, PAGE_SIZE, "0x%llx\n", entry->md.phys_addr);
> > +}
> > +
> > +static ssize_t virt_addr_show(struct efi_runtime_map_entry *entry, char *buf)
> > +{
> > + return snprintf(buf, PAGE_SIZE, "0x%llx\n", entry->md.virt_addr);
> > +}
> > +
> > +static ssize_t num_pages_show(struct efi_runtime_map_entry *entry, char *buf)
> > +{
> > + return snprintf(buf, PAGE_SIZE, "0x%llx\n", entry->md.num_pages);
> > +}
> > +
> > +static ssize_t attribute_show(struct efi_runtime_map_entry *entry, char *buf)
> > +{
> > + return snprintf(buf, PAGE_SIZE, "0x%llx\n", entry->md.attribute);
> > +}
>
> I can see the macro here writing itself too :)
Thanks for catching, actually I have changed it in my local tree since your
last comment about other file.
>
> > +
> > +static inline struct map_attribute *to_map_attr(struct attribute *attr)
> > +{
> > + return container_of(attr, struct map_attribute, attr);
> > +}
> > +
> > +static ssize_t map_attr_show(struct kobject *kobj,
> > + struct attribute *attr, char *buf)
> > +{
> > + struct efi_runtime_map_entry *entry = to_map_entry(kobj);
> > + struct map_attribute *map_attr = to_map_attr(attr);
> > +
> > + return map_attr->show(entry, buf);
> > +}
> > +
> > +static struct kset *map_kset;
> > +
> > +/*
> > + * Add map entry on sysfs
> > + */
>
> The function name should be enough - no need for the comment.
Ok.
>
> > +static struct efi_runtime_map_entry *add_sysfs_runtime_map_entry(int nr)
> > +{
> > + int ret;
> > + unsigned long desc_size;
> > + struct efi_runtime_map_entry *entry;
> > + struct efi_info *e = &boot_params.efi_info;
> > +
> > + if (!map_kset) {
> > + map_kset = kset_create_and_add("runtime-map", NULL,
> > + efi_kobj);
> > + if (!map_kset)
> > + return ERR_PTR(-ENOMEM);
> > + }
> > +
> > + entry = kzalloc(sizeof(*entry), GFP_KERNEL);
> > + if (!entry) {
> > + kset_unregister(map_kset);
> > + return entry;
> > + }
> > +
> > + desc_size = e->efi_memdesc_size;
> > + memcpy(&entry->md, efi_runtime_map + nr * desc_size,
> > + sizeof(efi_memory_desc_t));
>
> If you're going to break this line anyway, you don't really need the
> local desc_size variable:
Ok
>
> memcpy(&entry->md,
> efi_runtime_map + nr * e->efi_memdesc_size,
> sizeof(efi_memory_desc_t));
>
> > +
> > + kobject_init(&entry->kobj, &map_ktype);
> > + entry->kobj.kset = map_kset;
> > + ret = kobject_add(&entry->kobj, NULL, "%d", nr);
> > + if (ret) {
> > + kobject_put(&entry->kobj);
> > + kset_unregister(map_kset);
> > + return ERR_PTR(ret);
> > + }
> > +
> > + return entry;
> > +}
> > +
> > +static int __init efi_runtime_map_init(void)
> > +{
> > + int i, j, ret = 0;
> > + struct efi_runtime_map_entry *entry;
> > +
> > + if (!efi_runtime_map) {
> > + pr_warn("no efi_runtime_map found\n");
> > + return -EINVAL;
> > + }
> > +
> > + map_entries = kzalloc(nr_efi_runtime_map *
> > + sizeof(struct efi_runtime_map_entry *), GFP_KERNEL);
>
> Shorten:
>
> map_entries = kzalloc(nr_efi_runtime_map * sizeof(entry), GFP_KERNEL);
Will do.
>
> > + if (!map_entries)
> > + return -ENOMEM;
> > +
> > + for (i = 0; i < nr_efi_runtime_map; i++) {
> > + entry = add_sysfs_runtime_map_entry(i);
> > + if (IS_ERR(entry)) {
> > + for (j = i - 1; j > 0; j--) {
> > + entry = *(map_entries + j);
> > + kobject_put(&entry->kobj);
> > + }
> > + if (map_kset)
> > + kset_unregister(map_kset);
> > + ret = PTR_ERR(entry);
> > + goto out;
> > + }
>
> You can move this error unwinding path away from the main loop:
>
> if (IS_ERR(entry))
> goto out;
>
> and make it more readable.
Ok.
>
> > + *(map_entries + i) = entry;
> > + }
> > +
> > +out:
> > + return ret;
> > +}
> > +late_initcall(efi_runtime_map_init);
>
> Why an initcall? Can't we run this from efisubsys_init()?
Yes, we can, will do.
Thanks for review
Dave
