* [PATCH 0/2] kho: expose incoming sub-blob metadata instead of contents
@ 2026-11-17 9:29 Long Wei
2026-11-17 9:29 ` [PATCH 1/2] " Long Wei
` (2 more replies)
0 siblings, 3 replies; 11+ messages in thread
From: Long Wei @ 2026-11-17 9:29 UTC (permalink / raw)
To: rppt, pasha.tatashin, pratyush, akpm
Cc: graf, leitao, corbet, rdunlap, linux-mm, kexec, linux-doc,
linux-kernel, hewenliang4, Long Wei
Incoming sub-blob pages under kho/in/sub_fdts are owned by their consumers
and have independent lifetimes. Consumers restore and free them at their own
initcall level, which is always earlier than kho_init() (fs_initcall) where
the debugfs view is built. LUO, for example, frees its state in
early_initcall, so in/sub_fdts/LUO ends up pointing at a page that has
already been returned to the page allocator. Reading the file then triggers
a KASAN use-after-free in _copy_to_user(), with the backing page already
marked page_type: f0(buddy).
KHO does not own these pages, so it cannot safely expose their contents at
any point. Instead of trying to track which blobs are still alive, replace
the raw blob nodes with a read-only in/blobs file that reports the name,
physical address and size of every sub-blob referenced by the incoming FDT.
The FDT itself is reserved by KHO and remains available as in/fdt.
The first patch does the change, the second updates the documentation.
Long Wei (2):
kho: expose incoming sub-blob metadata instead of contents
docs: kho: describe the in/blobs debugfs file
Documentation/admin-guide/mm/kho.rst | 6 +--
kernel/liveupdate/kexec_handover_debugfs.c | 83 +++++++++++++++---------------
2 files changed, 45 insertions(+), 44 deletions(-)
--
2.43.0
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 1/2] kho: expose incoming sub-blob metadata instead of contents
2026-11-17 9:29 [PATCH 0/2] kho: expose incoming sub-blob metadata instead of contents Long Wei
@ 2026-11-17 9:29 ` Long Wei
2026-11-18 16:38 ` Pasha Tatashin
2026-11-17 9:29 ` [PATCH 2/2] docs: kho: describe the in/blobs debugfs file Long Wei
2026-11-19 6:37 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Long Wei
2 siblings, 1 reply; 11+ messages in thread
From: Long Wei @ 2026-11-17 9:29 UTC (permalink / raw)
To: rppt, pasha.tatashin, pratyush, akpm
Cc: graf, leitao, corbet, rdunlap, linux-mm, kexec, linux-doc,
linux-kernel, hewenliang4, Long Wei
Incoming sub-blob pages are owned by their consumers and have independent
lifetimes. Some consumers, such as LUO, restore and free their state in
early_initcall, i.e. before kho_init() (fs_initcall) builds the debugfs
view:
early_initcall: luo_early_startup()
-> kho_retrieve_subtree("LUO")
-> kho_restore_free(luo_ser) /* page is freed */
fs_initcall: kho_init()
-> kho_in_debugfs_init()
-> phys_to_virt(luo_ser_phys)
-> debugfs_create_blob("LUO", ...)
debugfs therefore ends up pointing at a page that was already handed back
to the page allocator, and reading the file triggers a use-after-free:
# cat /sys/kernel/debug/kho/in/sub_fdts/LUO
==================================================================
BUG: KASAN: use-after-free in _copy_to_user+0x37/0x60
Read of size 32 at addr ffff88810309c000 by task cat/27436
Call Trace:
<TASK>
kasan_report+0xb9/0xf0
kasan_check_range+0x100/0x1b0
_copy_to_user+0x37/0x60
simple_read_from_buffer+0x77/0xd0
read_file_blob+0x75/0xa0
vfs_read+0x12d/0x520
ksys_read+0xc3/0x160
do_syscall_64+0x97/0x4b0
entry_SYSCALL_64_after_hwframe+0x76/0x7e
</TASK>
The buggy address belongs to the physical page:
page: refcount:0 mapcount:0 mapping:0000000000000000 index:0x1fd
pfn:0x10309c
flags: 0x17ffffc0000000(node=0|zone=2|lastcpupid=0x1fffff)
page_type: f0(buddy)
==================================================================
The page is already in the buddy allocator (page_type: f0(buddy),
refcount 0), which confirms it was freed before the debugfs view was
created.
KHO does not own these pages, so it cannot safely expose their contents at
all. Drop the in/sub_fdts raw blob nodes and replace them with a
read-only in/blobs metadata file that lists the name, physical address and
size of every incoming sub-blob. The FDT itself is reserved by KHO and is
still exposed as in/fdt.
Fixes: 062dd306d99c ("kho: fix kho_in_debugfs_init() to handle non-FDT blobs")
Signed-off-by: Long Wei <longwei27@huawei.com>
---
kernel/liveupdate/kexec_handover_debugfs.c | 83 +++++++++++-----------
1 file changed, 42 insertions(+), 41 deletions(-)
diff --git a/kernel/liveupdate/kexec_handover_debugfs.c b/kernel/liveupdate/kexec_handover_debugfs.c
index 7d74477da0b7..3624d4bf44b2 100644
--- a/kernel/liveupdate/kexec_handover_debugfs.c
+++ b/kernel/liveupdate/kexec_handover_debugfs.c
@@ -95,67 +95,68 @@ static int bootmem_len_show(struct seq_file *m, void *v)
}
DEFINE_SHOW_ATTRIBUTE(bootmem_len);
-__init void kho_in_debugfs_init(struct kho_debugfs *dbg, const void *fdt)
+static int kho_in_blobs_show(struct seq_file *m, void *v)
{
- struct dentry *dir, *sub_fdt_dir;
- int err, child;
-
- INIT_LIST_HEAD(&dbg->fdt_list);
-
- dir = debugfs_create_dir("in", debugfs_root);
- if (IS_ERR(dir)) {
- err = PTR_ERR(dir);
- goto err_out;
- }
+ const void *fdt = m->private;
+ int child, len;
- sub_fdt_dir = debugfs_create_dir("sub_fdts", dir);
- if (IS_ERR(sub_fdt_dir)) {
- err = PTR_ERR(sub_fdt_dir);
- goto err_rmdir;
+ if (fdt_check_header(fdt)) {
+ seq_puts(m, "invalid incoming FDT\n");
+ return 0;
}
- err = __kho_debugfs_blob_add(&dbg->fdt_list, dir, "fdt", fdt,
- fdt_totalsize(fdt));
- if (err)
- goto err_rmdir;
-
+ seq_puts(m, "# name pa size\n");
fdt_for_each_subnode(child, fdt, 0) {
- int len = 0;
const char *name = fdt_get_name(fdt, child, NULL);
const u64 *blob_phys;
const u64 *blob_size;
- void *blob;
blob_phys = fdt_getprop(fdt, child,
KHO_SUB_TREE_PROP_NAME, &len);
- if (!blob_phys)
- continue;
- if (len != sizeof(*blob_phys)) {
- pr_warn("node %s prop %s has invalid length: %d\n",
- name, KHO_SUB_TREE_PROP_NAME, len);
+ if (!blob_phys || len != sizeof(*blob_phys))
continue;
- }
blob_size = fdt_getprop(fdt, child,
KHO_SUB_TREE_SIZE_PROP_NAME, &len);
- if (!blob_size || len != sizeof(*blob_size)) {
- pr_warn("node %s missing or invalid %s property\n",
- name, KHO_SUB_TREE_SIZE_PROP_NAME);
+ if (!blob_size || len != sizeof(*blob_size))
continue;
- }
- blob = phys_to_virt(*blob_phys);
- err = __kho_debugfs_blob_add(&dbg->fdt_list, sub_fdt_dir, name,
- blob, *blob_size);
- if (err) {
- pr_warn("failed to add blob %s to debugfs: %pe\n",
- name, ERR_PTR(err));
- continue;
- }
+ seq_printf(m, "%s 0x%016llx %llu\n", name,
+ (u64)*blob_phys, (u64)*blob_size);
}
+ return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(kho_in_blobs);
+
+/*
+ * Show the metadata of the sub-blobs referenced by the incoming FDT. Each
+ * line contains the name, physical address and size of a sub-blob. The raw
+ * contents of the sub-blobs are not exposed.
+ */
+__init void kho_in_debugfs_init(struct kho_debugfs *dbg, const void *fdt)
+{
+ struct dentry *dir;
+ int err;
+
+ INIT_LIST_HEAD(&dbg->fdt_list);
+
+ dir = debugfs_create_dir("in", debugfs_root);
+ if (IS_ERR(dir)) {
+ err = PTR_ERR(dir);
+ goto err_out;
+ }
+
+ err = __kho_debugfs_blob_add(&dbg->fdt_list, dir, "fdt", fdt,
+ fdt_totalsize(fdt));
+ if (err)
+ goto err_rmdir;
+
+ debugfs_create_file("blobs", 0400, dir, (void *)fdt,
+ &kho_in_blobs_fops);
+
dbg->dir = dir;
- dbg->sub_fdt_dir = sub_fdt_dir;
+ dbg->sub_fdt_dir = NULL;
return;
err_rmdir:
--
2.43.0
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH 2/2] docs: kho: describe the in/blobs debugfs file
2026-11-17 9:29 [PATCH 0/2] kho: expose incoming sub-blob metadata instead of contents Long Wei
2026-11-17 9:29 ` [PATCH 1/2] " Long Wei
@ 2026-11-17 9:29 ` Long Wei
2026-11-19 6:37 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Long Wei
2 siblings, 0 replies; 11+ messages in thread
From: Long Wei @ 2026-11-17 9:29 UTC (permalink / raw)
To: rppt, pasha.tatashin, pratyush, akpm
Cc: graf, leitao, corbet, rdunlap, linux-mm, kexec, linux-doc,
linux-kernel, hewenliang4, Long Wei
The incoming sub-blob raw nodes under kho/in/sub_fdts no longer exist:
the sub-blob pages are owned by their consumers and may already be freed
by the time the debugfs view is created, so exposing them was unsafe.
Document the in/blobs metadata file that replaces them.
Signed-off-by: Long Wei <longwei27@huawei.com>
---
Documentation/admin-guide/mm/kho.rst | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/Documentation/admin-guide/mm/kho.rst b/Documentation/admin-guide/mm/kho.rst
index 902d171a780a..602c8ce90e9a 100644
--- a/Documentation/admin-guide/mm/kho.rst
+++ b/Documentation/admin-guide/mm/kho.rst
@@ -118,6 +118,6 @@ stabilized.
device tree. This file may disappear when all consumers of
it finished to interpret their metadata.
-``/sys/kernel/debug/kho/in/sub_fdts/``
- Similar to ``kho/out/sub_fdts/``, but contains sub blobs
- of KHO producers passed from the old kernel.
+``/sys/kernel/debug/kho/in/blobs``
+ Metadata for the sub-blobs referenced by the incoming FDT. Each line
+ contains the name, physical address and size of a sub-blob.
--
2.43.0
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH 1/2] kho: expose incoming sub-blob metadata instead of contents
2026-11-17 9:29 ` [PATCH 1/2] " Long Wei
@ 2026-11-18 16:38 ` Pasha Tatashin
0 siblings, 0 replies; 11+ messages in thread
From: Pasha Tatashin @ 2026-11-18 16:38 UTC (permalink / raw)
To: Long Wei
Cc: rppt, pasha.tatashin, pratyush, akpm, graf, leitao, corbet,
rdunlap, linux-mm, kexec, linux-doc, linux-kernel, hewenliang4
> Incoming sub-blob pages are owned by their consumers and have independent
> lifetimes. Some consumers, such as LUO, restore and free their state in
> early_initcall, i.e. before kho_init() (fs_initcall) builds the debugfs
> view:
>
> early_initcall: luo_early_startup()
> -> kho_retrieve_subtree("LUO")
> -> kho_restore_free(luo_ser) /* page is freed */
>
> fs_initcall: kho_init()
> -> kho_in_debugfs_init()
> -> phys_to_virt(luo_ser_phys)
> -> debugfs_create_blob("LUO", ...)
>
> debugfs therefore ends up pointing at a page that was already handed back
> to the page allocator, and reading the file triggers a use-after-free:
>
> # cat /sys/kernel/debug/kho/in/sub_fdts/LUO
> ==================================================================
> BUG: KASAN: use-after-free in _copy_to_user+0x37/0x60
> Read of size 32 at addr ffff88810309c000 by task cat/27436
>
> Call Trace:
> <TASK>
> kasan_report+0xb9/0xf0
> kasan_check_range+0x100/0x1b0
> _copy_to_user+0x37/0x60
> simple_read_from_buffer+0x77/0xd0
> read_file_blob+0x75/0xa0
> vfs_read+0x12d/0x520
> ksys_read+0xc3/0x160
> do_syscall_64+0x97/0x4b0
> entry_SYSCALL_64_after_hwframe+0x76/0x7e
> </TASK>
>
> The buggy address belongs to the physical page:
> page: refcount:0 mapcount:0 mapping:0000000000000000 index:0x1fd
> pfn:0x10309c
> flags: 0x17ffffc0000000(node=0|zone=2|lastcpupid=0x1fffff)
> page_type: f0(buddy)
> ==================================================================
>
> The page is already in the buddy allocator (page_type: f0(buddy),
> refcount 0), which confirms it was freed before the debugfs view was
> created.
>
> KHO does not own these pages, so it cannot safely expose their contents at
> all. Drop the in/sub_fdts raw blob nodes and replace them with a
> read-only in/blobs metadata file that lists the name, physical address and
> size of every incoming sub-blob. The FDT itself is reserved by KHO and is
> still exposed as in/fdt.
I suggest we simply remove in/sub_fdts without adding in/blobs. All of
the metadata you are proposing to expose via in/blobs is already present
in in/fdt (and can be parsed from userspace if needed), so I do not
think it warrants adding another kernel debugfs interface. Keeping this
to just removing in/sub_fdts (and updating the doc) also keeps the fix
minimal for -rc.
>
> Fixes: 062dd306d99c ("kho: fix kho_in_debugfs_init() to handle non-FDT blobs")
This UAF has been there since in/sub_fdts was first added, so I think
this should be: Fixes: c609c144b0e8 ("kexec: add KHO parsing support")
--
Pasha
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view
2026-11-17 9:29 [PATCH 0/2] kho: expose incoming sub-blob metadata instead of contents Long Wei
2026-11-17 9:29 ` [PATCH 1/2] " Long Wei
2026-11-17 9:29 ` [PATCH 2/2] docs: kho: describe the in/blobs debugfs file Long Wei
@ 2026-11-19 6:37 ` Long Wei
2026-11-19 6:37 ` [PATCH v2 1/2] kho: do not expose incoming sub-blob contents in debugfs Long Wei
` (3 more replies)
2 siblings, 4 replies; 11+ messages in thread
From: Long Wei @ 2026-11-19 6:37 UTC (permalink / raw)
To: rppt, pasha.tatashin, pratyush, akpm
Cc: graf, leitao, corbet, rdunlap, linux-mm, kexec, linux-doc,
linux-kernel, hewenliang4, Long Wei
kho/in/sub_fdts exposes the raw contents of the sub-blobs described by
the incoming KHO FDT. KHO does not own those pages: they belong to the
producers' consumers and have independent lifetimes. LUO, for example,
restores and frees its state in early_initcall, before kho_init()
(fs_initcall) creates the debugfs view, so the blobs point at pages that
are already back in the buddy allocator. Reading them triggers a KASAN
use-after-free in _copy_to_user().
The same information is already available to user space from in/fdt,
which KHO does own and which stays reserved. Drop the directory rather
than adding another interface to describe it, keeping the -rc fix small.
Patch 1 removes the interface, patch 2 drops its documentation.
Changes in v2:
- Drop the in/blobs metadata file entirely
- Just remove in/sub_fdts to fix the UAF, no replacement interface
- Update Fixes tag to c609c144b0e8 (the commit that first added sub_fdts)
- Remove the now-unused #include <linux/io.h>
Long Wei (2):
kho: do not expose incoming sub-blob contents in debugfs
docs: kho: drop kho/in/sub_fdts description
Documentation/admin-guide/mm/kho.rst | 4 --
kernel/liveupdate/kexec_handover_debugfs.c | 47 +---------------------
2 files changed, 2 insertions(+), 49 deletions(-)
--
2.43.0
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH v2 1/2] kho: do not expose incoming sub-blob contents in debugfs
2026-11-19 6:37 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Long Wei
@ 2026-11-19 6:37 ` Long Wei
2026-11-19 8:26 ` Mike Rapoport
2026-11-19 6:38 ` [PATCH v2 2/2] docs: kho: drop kho/in/sub_fdts description Long Wei
` (2 subsequent siblings)
3 siblings, 1 reply; 11+ messages in thread
From: Long Wei @ 2026-11-19 6:37 UTC (permalink / raw)
To: rppt, pasha.tatashin, pratyush, akpm
Cc: graf, leitao, corbet, rdunlap, linux-mm, kexec, linux-doc,
linux-kernel, hewenliang4, Long Wei
Incoming sub-blob pages are owned by their consumers and have independent
lifetimes. Some consumers, such as LUO, restore and free their state in
early_initcall, i.e. before kho_init() (fs_initcall) builds the debugfs
view:
early_initcall: luo_early_startup()
-> kho_retrieve_subtree("LUO")
-> kho_restore_free(luo_ser) /* page is freed */
fs_initcall: kho_init()
-> kho_in_debugfs_init()
-> phys_to_virt(luo_ser_phys)
-> debugfs_create_blob("LUO", ...)
debugfs therefore ends up pointing at a page that was already handed back
to the page allocator, and reading the file triggers a use-after-free:
# cat /sys/kernel/debug/kho/in/sub_fdts/LUO
==================================================================
BUG: KASAN: use-after-free in _copy_to_user+0x37/0x60
Read of size 32 at addr ffff88810309c000 by task cat/27436
Call Trace:
<TASK>
kasan_report+0xb9/0xf0
kasan_check_range+0x100/0x1b0
_copy_to_user+0x37/0x60
simple_read_from_buffer+0x77/0xd0
read_file_blob+0x75/0xa0
vfs_read+0x12d/0x520
ksys_read+0xc3/0x160
do_syscall_64+0x97/0x4b0
entry_SYSCALL_64_after_hwframe+0x76/0x7e
</TASK>
The buggy address belongs to the physical page:
page: refcount:0 mapcount:0 mapping:0000000000000000 index:0x1fd
pfn:0x10309c
flags: 0x17ffffc0000000(node=0|zone=2|lastcpupid=0x1fffff)
page_type: f0(buddy)
==================================================================
The page is already in the buddy allocator (page_type: f0(buddy),
refcount 0), which confirms it was freed before the debugfs view was
created.
KHO does not own these pages, so it cannot safely expose their contents at
all. Drop the in/sub_fdts directory. The incoming FDT is reserved by KHO
and remains exposed as in/fdt, where the same metadata is available.
Fixes: c609c144b0e8 ("kexec: add KHO parsing support")
Signed-off-by: Long Wei <longwei27@huawei.com>
---
kernel/liveupdate/kexec_handover_debugfs.c | 47 +---------------------
1 file changed, 2 insertions(+), 45 deletions(-)
diff --git a/kernel/liveupdate/kexec_handover_debugfs.c b/kernel/liveupdate/kexec_handover_debugfs.c
index 7d74477da0b7..3a91c47ad7a6 100644
--- a/kernel/liveupdate/kexec_handover_debugfs.c
+++ b/kernel/liveupdate/kexec_handover_debugfs.c
@@ -10,7 +10,6 @@
#define pr_fmt(fmt) "KHO: " fmt
#include <linux/init.h>
-#include <linux/io.h>
#include <linux/libfdt.h>
#include <linux/mm.h>
#include <linux/kho/abi/kexec_handover.h>
@@ -97,8 +96,8 @@ DEFINE_SHOW_ATTRIBUTE(bootmem_len);
__init void kho_in_debugfs_init(struct kho_debugfs *dbg, const void *fdt)
{
- struct dentry *dir, *sub_fdt_dir;
- int err, child;
+ struct dentry *dir;
+ int err;
INIT_LIST_HEAD(&dbg->fdt_list);
@@ -108,54 +107,12 @@ __init void kho_in_debugfs_init(struct kho_debugfs *dbg, const void *fdt)
goto err_out;
}
- sub_fdt_dir = debugfs_create_dir("sub_fdts", dir);
- if (IS_ERR(sub_fdt_dir)) {
- err = PTR_ERR(sub_fdt_dir);
- goto err_rmdir;
- }
-
err = __kho_debugfs_blob_add(&dbg->fdt_list, dir, "fdt", fdt,
fdt_totalsize(fdt));
if (err)
goto err_rmdir;
- fdt_for_each_subnode(child, fdt, 0) {
- int len = 0;
- const char *name = fdt_get_name(fdt, child, NULL);
- const u64 *blob_phys;
- const u64 *blob_size;
- void *blob;
-
- blob_phys = fdt_getprop(fdt, child,
- KHO_SUB_TREE_PROP_NAME, &len);
- if (!blob_phys)
- continue;
- if (len != sizeof(*blob_phys)) {
- pr_warn("node %s prop %s has invalid length: %d\n",
- name, KHO_SUB_TREE_PROP_NAME, len);
- continue;
- }
-
- blob_size = fdt_getprop(fdt, child,
- KHO_SUB_TREE_SIZE_PROP_NAME, &len);
- if (!blob_size || len != sizeof(*blob_size)) {
- pr_warn("node %s missing or invalid %s property\n",
- name, KHO_SUB_TREE_SIZE_PROP_NAME);
- continue;
- }
-
- blob = phys_to_virt(*blob_phys);
- err = __kho_debugfs_blob_add(&dbg->fdt_list, sub_fdt_dir, name,
- blob, *blob_size);
- if (err) {
- pr_warn("failed to add blob %s to debugfs: %pe\n",
- name, ERR_PTR(err));
- continue;
- }
- }
-
dbg->dir = dir;
- dbg->sub_fdt_dir = sub_fdt_dir;
return;
err_rmdir:
--
2.43.0
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH v2 2/2] docs: kho: drop kho/in/sub_fdts description
2026-11-19 6:37 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Long Wei
2026-11-19 6:37 ` [PATCH v2 1/2] kho: do not expose incoming sub-blob contents in debugfs Long Wei
@ 2026-11-19 6:38 ` Long Wei
2026-11-19 6:53 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Pasha Tatashin
2026-11-19 8:26 ` Mike Rapoport
3 siblings, 0 replies; 11+ messages in thread
From: Long Wei @ 2026-11-19 6:38 UTC (permalink / raw)
To: rppt, pasha.tatashin, pratyush, akpm
Cc: graf, leitao, corbet, rdunlap, linux-mm, kexec, linux-doc,
linux-kernel, hewenliang4, Long Wei
The incoming sub-blob debugfs directory is gone, so remove its description
from the KHO admin guide.
Signed-off-by: Long Wei <longwei27@huawei.com>
---
Documentation/admin-guide/mm/kho.rst | 4 ----
1 file changed, 4 deletions(-)
diff --git a/Documentation/admin-guide/mm/kho.rst b/Documentation/admin-guide/mm/kho.rst
index 902d171a780a..4281da6be5fa 100644
--- a/Documentation/admin-guide/mm/kho.rst
+++ b/Documentation/admin-guide/mm/kho.rst
@@ -117,7 +117,3 @@ stabilized.
kernel's state is in this file in the format of flattened
device tree. This file may disappear when all consumers of
it finished to interpret their metadata.
-
-``/sys/kernel/debug/kho/in/sub_fdts/``
- Similar to ``kho/out/sub_fdts/``, but contains sub blobs
- of KHO producers passed from the old kernel.
--
2.43.0
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view
2026-11-19 6:37 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Long Wei
2026-11-19 6:37 ` [PATCH v2 1/2] kho: do not expose incoming sub-blob contents in debugfs Long Wei
2026-11-19 6:38 ` [PATCH v2 2/2] docs: kho: drop kho/in/sub_fdts description Long Wei
@ 2026-11-19 6:53 ` Pasha Tatashin
2026-11-19 8:26 ` Mike Rapoport
3 siblings, 0 replies; 11+ messages in thread
From: Pasha Tatashin @ 2026-11-19 6:53 UTC (permalink / raw)
To: Long Wei
Cc: rppt, pasha.tatashin, pratyush, akpm, graf, leitao, corbet,
rdunlap, linux-mm, kexec, linux-doc, linux-kernel, hewenliang4
On Thu, 19 Nov 2026 14:37:58 +0800, Long Wei <longwei27@huawei.com> wrote:
> kho: drop the incoming sub-blob debugfs view
>
> kho/in/sub_fdts exposes the raw contents of the sub-blobs described by
> the incoming KHO FDT. KHO does not own those pages: they belong to the
> producers' consumers and have independent lifetimes. LUO, for example,
> restores and frees its state in early_initcall, before kho_init()
> (fs_initcall) creates the debugfs view, so the blobs point at pages that
> are already back in the buddy allocator. Reading them triggers a KASAN
> use-after-free in _copy_to_user().
>
> [...]
Acked-by: Pasha Tatashin <pasha.tatashin@soleen.com>
--
Pasha
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view
2026-11-19 6:37 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Long Wei
` (2 preceding siblings ...)
2026-11-19 6:53 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Pasha Tatashin
@ 2026-11-19 8:26 ` Mike Rapoport
2026-11-19 9:11 ` longwei (I)
3 siblings, 1 reply; 11+ messages in thread
From: Mike Rapoport @ 2026-11-19 8:26 UTC (permalink / raw)
To: Long Wei
Cc: rppt, pasha.tatashin, pratyush, akpm, graf, leitao, corbet,
rdunlap, linux-mm, kexec, linux-doc, linux-kernel, hewenliang4
Hi,
Please don't send vN+1 as a reply to vN.
> kho/in/sub_fdts exposes the raw contents of the sub-blobs described by
> the incoming KHO FDT. KHO does not own those pages: they belong to the
> producers' consumers and have independent lifetimes. LUO, for example,
> restores and frees its state in early_initcall, before kho_init()
> (fs_initcall) creates the debugfs view, so the blobs point at pages that
> are already back in the buddy allocator. Reading them triggers a KASAN
> use-after-free in _copy_to_user().
>
> The same information is already available to user space from in/fdt,
> which KHO does own and which stays reserved. Drop the directory rather
> than adding another interface to describe it, keeping the -rc fix small.
>
> Patch 1 removes the interface, patch 2 drops its documentation.
Patch 2 is only four lines of documentation for the interface patch 1
removes. It could be folded into patch 1 so that a stable backport
removes the documentation together with the code.
--
Sincerely yours,
Mike.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v2 1/2] kho: do not expose incoming sub-blob contents in debugfs
2026-11-19 6:37 ` [PATCH v2 1/2] kho: do not expose incoming sub-blob contents in debugfs Long Wei
@ 2026-11-19 8:26 ` Mike Rapoport
0 siblings, 0 replies; 11+ messages in thread
From: Mike Rapoport @ 2026-11-19 8:26 UTC (permalink / raw)
To: Long Wei
Cc: rppt, pasha.tatashin, pratyush, akpm, graf, leitao, corbet,
rdunlap, linux-mm, kexec, linux-doc, linux-kernel, hewenliang4
Hi,
> Incoming sub-blob pages are owned by their consumers and have independent
> lifetimes. Some consumers, such as LUO, restore and free their state in
> early_initcall, i.e. before kho_init() (fs_initcall) builds the debugfs
> view:
>
> early_initcall: luo_early_startup()
> -> kho_retrieve_subtree("LUO")
> -> kho_restore_free(luo_ser) /* page is freed */
>
> fs_initcall: kho_init()
> -> kho_in_debugfs_init()
> -> phys_to_virt(luo_ser_phys)
> -> debugfs_create_blob("LUO", ...)
>
> debugfs therefore ends up pointing at a page that was already handed back
> to the page allocator, and reading the file triggers a use-after-free:
>
> # cat /sys/kernel/debug/kho/in/sub_fdts/LUO
> ==================================================================
> BUG: KASAN: use-after-free in _copy_to_user+0x37/0x60
> Read of size 32 at addr ffff88810309c000 by task cat/27436
>
> Call Trace:
> <TASK>
> kasan_report+0xb9/0xf0
> kasan_check_range+0x100/0x1b0
> _copy_to_user+0x37/0x60
> simple_read_from_buffer+0x77/0xd0
> read_file_blob+0x75/0xa0
> vfs_read+0x12d/0x520
> ksys_read+0xc3/0x160
> do_syscall_64+0x97/0x4b0
> entry_SYSCALL_64_after_hwframe+0x76/0x7e
> </TASK>
>
> The buggy address belongs to the physical page:
> page: refcount:0 mapcount:0 mapping:0000000000000000 index:0x1fd
> pfn:0x10309c
> flags: 0x17ffffc0000000(node=0|zone=2|lastcpupid=0x1fffff)
> page_type: f0(buddy)
> ==================================================================
>
> The page is already in the buddy allocator (page_type: f0(buddy),
> refcount 0), which confirms it was freed before the debugfs view was
> created.
>
> KHO does not own these pages, so it cannot safely expose their contents at
> all. Drop the in/sub_fdts directory. The incoming FDT is reserved by KHO
> and remains exposed as in/fdt, where the same metadata is available.
>
> Fixes: c609c144b0e8 ("kexec: add KHO parsing support")
> Signed-off-by: Long Wei <longwei27@huawei.com>
The change is fine, but the Fixes: tag should name the commit that
made the use-after-free reachable.
c609c144b0e8 ("kexec: add KHO parsing support") added in/sub_fdts, but
only memblock consumed an incoming sub-tree back then, and it never
freed it. Preserved pages that nobody restores stay reserved, so the
debugfs blobs pointed at valid memory. The same was true for test_kho
and for LUO while it kept luo_global.fdt_in.
The page is first freed early in boot by d376e4b55c9a0 ("liveupdate:
centralize state management into struct luo_ser"). That commit added
kho_restore_free(luo_ser) to luo_early_startup(), and it is the path
in the changelog and the KASAN report. It went into v7.2-rc1.
Pasha asked for c609c144b0e8 in v1 because in/sub_fdts never owned
those pages. That is a fair point about the design, but with that tag
stable will try to backport the patch to every tree since v6.16. There
is no bug in those trees, and the code there lives in
kernel/kexec_handover.c, so the patch won't apply. I'd use
Fixes: d376e4b55c9a0 ("liveupdate: centralize state management into struct luo_ser")
and say in the changelog that in/sub_fdts has exposed memory KHO does not
own since c609c144b0e8.
--
Sincerely yours,
Mike.
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view
2026-11-19 8:26 ` Mike Rapoport
@ 2026-11-19 9:11 ` longwei (I)
0 siblings, 0 replies; 11+ messages in thread
From: longwei (I) @ 2026-11-19 9:11 UTC (permalink / raw)
To: Mike Rapoport
Cc: pasha.tatashin, pratyush, akpm, graf, leitao, corbet, rdunlap,
linux-mm, kexec, linux-doc, linux-kernel, hewenliang4
Hi Mike,
Thanks for the detailed explanation! I now understand why c609c144b0e8
is not the right Fixes target — the pages weren't freed back then, so
there was no UAF. The kho_restore_free() path in d376e4b55c9a0 is what
actually made it reachable.
I will send v3 with the following changes:
- Fold patch 2 (documentation removal) into patch 1, so stable
backports pick up both code and docs together
- Update Fixes tag to d376e4b55c9a0 ("liveupdate: centralize state
management into struct luo_ser")
Best,
Long Wei
在 2026/11/19 16:26, Mike Rapoport 写道:
> Hi,
>
> Please don't send vN+1 as a reply to vN.
>
>> kho/in/sub_fdts exposes the raw contents of the sub-blobs described by
>> the incoming KHO FDT. KHO does not own those pages: they belong to the
>> producers' consumers and have independent lifetimes. LUO, for example,
>> restores and frees its state in early_initcall, before kho_init()
>> (fs_initcall) creates the debugfs view, so the blobs point at pages that
>> are already back in the buddy allocator. Reading them triggers a KASAN
>> use-after-free in _copy_to_user().
>>
>> The same information is already available to user space from in/fdt,
>> which KHO does own and which stays reserved. Drop the directory rather
>> than adding another interface to describe it, keeping the -rc fix small.
>>
>> Patch 1 removes the interface, patch 2 drops its documentation.
>
> Patch 2 is only four lines of documentation for the interface patch 1
> removes. It could be folded into patch 1 so that a stable backport
> removes the documentation together with the code.
>
^ permalink raw reply [flat|nested] 11+ messages in thread
end of thread, other threads:[~2026-11-19 9:13 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-11-17 9:29 [PATCH 0/2] kho: expose incoming sub-blob metadata instead of contents Long Wei
2026-11-17 9:29 ` [PATCH 1/2] " Long Wei
2026-11-18 16:38 ` Pasha Tatashin
2026-11-17 9:29 ` [PATCH 2/2] docs: kho: describe the in/blobs debugfs file Long Wei
2026-11-19 6:37 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Long Wei
2026-11-19 6:37 ` [PATCH v2 1/2] kho: do not expose incoming sub-blob contents in debugfs Long Wei
2026-11-19 8:26 ` Mike Rapoport
2026-11-19 6:38 ` [PATCH v2 2/2] docs: kho: drop kho/in/sub_fdts description Long Wei
2026-11-19 6:53 ` [PATCH v2 0/2] kho: drop the incoming sub-blob debugfs view Pasha Tatashin
2026-11-19 8:26 ` Mike Rapoport
2026-11-19 9:11 ` longwei (I)
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®