x86/kexec: fix memory leak of elf header buffer
Subsystems (5)
v2 cover x86/kexec: fix memory leak of elf header buffer
The memory leak is reported by kmemleak detector, has been existing for very long time. It casue much memory loss on large machine with huge memory hotplug which will trigger kdump kernel reloading many times, with kexec_file_load interface.
And in patch 2, clean up is done to remove unneeded arch_kexec_kernel_image_load() and rename functions.
V1 post can be found here:
https://lore.kernel.org/all/20211029072424.9109-1-bhe at redhat.com/T/#u
v1->v2:
No code change. The v1 post has been acked by Dave Young but not
merged. In v2, just merging the old patch 1 and 2 as per Dave's concern in v1 post.
Baoquan He (2):
x86/kexec: fix memory leak of elf header buffer
kexec_file: clean up arch_kexec_kernel_image_load
arch/x86/kernel/machine_kexec_64.c | 23 +++++++++-------------- include/linux/kexec.h | 1 - kernel/kexec_file.c | 9 ++------- 3 files changed, 11 insertions(+), 22 deletions(-)
Signature · 1 line
2.31.1
PATCH v2 1/2 x86/kexec: fix memory leak of elf header buffer
This is reported by kmemleak detector:
unreferenced object 0xffffc900002a9000 (size 4096):
comm "kexec", pid 14950, jiffies 4295110793 (age 373.951s) hex dump (first 32 bytes):
7f 45 4c 46 02 01 01 00 00 00 00 00 00 00 00 00 .ELF............
04 00 3e 00 01 00 00 00 00 00 00 00 00 00 00 00 ..>.............
backtrace:
[<0000000016a8ef9f>] __vmalloc_node_range+0x101/0x170
[<000000002b66b6c0>] __vmalloc_node+0xb4/0x160
[<00000000ad40107d>] crash_prepare_elf64_headers+0x8e/0xcd0
[<0000000019afff23>] crash_load_segments+0x260/0x470
[<0000000019ebe95c>] bzImage64_load+0x814/0xad0
[<0000000093e16b05>] arch_kexec_kernel_image_load+0x1be/0x2a0
[<000000009ef2fc88>] kimage_file_alloc_init+0x2ec/0x5a0
[<0000000038f5a97a>] __do_sys_kexec_file_load+0x28d/0x530
[<0000000087c19992>] do_syscall_64+0x3b/0x90
[<0000000066e063a4>] entry_SYSCALL_64_after_hwframe+0x44/0xaeIn crash_prepare_elf64_headers(), a buffer is allocated via vmalloc() to store elf headers. While it's not freed back to system correctly when kdump kernel is reloaded or unloaded. Then memory leak is caused. Fix it by introducing x86 specific function arch_kimage_file_post_load_cleanup(), and freeing the buffer there.
And also remove the incorrect elf header buffer freeing code. Before calling arch specific kexec_file loading function, the image instance has been initialized. So 'image->elf_headers' must be NULL. It doesn't make sense to free the elf header buffer in the place.
--- arch/x86/kernel/machine_kexec_64.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-)
arch/x86/kernel/machine_kexec_64.c +9 -3
@@ -374,9 +374,6 @@ void machine_kexec(struct kimage *image)#ifdef CONFIG_KEXEC_FILE void *arch_kexec_kernel_image_load (struct kimage *image) { -vfree (image->elf_headers); - image->elf_headers =NULL ; -if (!image->fops || !image->fops->load)return ERR_PTR (-ENOEXEC );
@@ -512,6 +509,15 @@ int arch_kexec_apply_relocations_add(struct purgatory_info *pi, (int)ELF64_R_TYPE(rel[i].r_info), value);return -ENOEXEC ; } + +int arch_kimage_file_post_load_cleanup (struct kimage *image) +{ +vfree (image->elf_headers); + image->elf_headers =NULL ; + image->elf_headers_sz =0 ; + +return kexec_image_post_load_cleanup_default (image); +}#endif /* CONFIG_KEXEC_FILE */ static int
Signature · 1 line
2.31.1
PATCH v2 2/2 kexec_file: clean up arch_kexec_kernel_image_load
In reply to Baoquan HeFunction arch_kexec_kernel_image_load() has a common weak version which is a wrapper of kexec_image_load_default() , and has only one arch dependent version in x86_64. Now the x86_64 dependent function is not needed any more. So clean it up.
And also rename kexec_image_load_default() to kexec_kernel_image_load() for better reflecting its functionality.
--- arch/x86/kernel/machine_kexec_64.c | 11 ----------- include/linux/kexec.h | 1 - kernel/kexec_file.c | 9 ++------- 3 files changed, 2 insertions(+), 19 deletions(-)
arch/x86/kernel/machine_kexec_64.c +0 -11
@@ -372,17 +372,6 @@ void machine_kexec(struct kimage *image)/* arch-dependent functionality related to kexec file-based syscall */ #ifdef CONFIG_KEXEC_FILE -void *arch_kexec_kernel_image_load (struct kimage *image) -{ -if (!image->fops || !image->fops->load) -return ERR_PTR (-ENOEXEC ); - -return image->fops->load (image, image->kernel_buf, - image->kernel_buf_len, image->initrd_buf, - image->initrd_buf_len, image->cmdline_buf, - image->cmdline_buf_len); -} -/* * Apply purgatory relocations. *
include/linux/kexec.h +0 -1
@@ -186,7 +186,6 @@ void *kexec_purgatory_get_symbol_addr(struct kimage *image, const char *name);/* Architectures may override the below functions */ int arch_kexec_kernel_image_probe (struct kimage *image,void *buf,unsigned long buf_len); -void *arch_kexec_kernel_image_load (struct kimage *image); int arch_kexec_apply_relocations_add(struct purgatory_info *pi, Elf_Shdr *section, const Elf_Shdr *relsec,
kernel/kexec_file.c +2 -7
@@ -60,7 +60,7 @@ int __weak arch_kexec_kernel_image_probe(struct kimage *image, void *buf,return kexec_image_probe_default (image, buf, buf_len); } -static void *kexec_image_load_default (struct kimage *image) +static void *kexec_kernel_image_load (struct kimage *image) {if (!image->fops || !image->fops->load)return ERR_PTR (-ENOEXEC );
@@ -71,11 +71,6 @@ static void *kexec_image_load_default(struct kimage *image) image->cmdline_buf_len); } -void *__weak arch_kexec_kernel_image_load (struct kimage *image) -{ -return kexec_image_load_default (image); -} -int kexec_image_post_load_cleanup_default (struct kimage *image) {if (!image->fops || !image->fops->cleanup)
@@ -279,7 +274,7 @@ kimage_file_prepare_segments(struct kimage *image, int kernel_fd, int initrd_fd,ima_add_kexec_buffer (image);/* Call arch image load handlers */ - ldata =arch_kexec_kernel_image_load (image); + ldata =kexec_kernel_image_load (image);if (IS_ERR (ldata)) { ret =PTR_ERR (ldata);
Signature · 1 line
2.31.1
Hi Andrew,
Quoted 4 lines · Baoquan He
The memory leak is reported by kmemleak detector, has been existing for very long time. It casue much memory loss on large machine with huge memory hotplug which will trigger kdump kernel reloading many times, with
kexec_file_loadinterface.
Could you merge these two patches? Or should I ping x86 maintainers to take them?
Thanks
Baoquan
Quoted 24 lines
And in patch 2, clean up is done to remove unneeded
arch_kexec_kernel_image_load()and rename functions.V1 post can be found here:
https://lore.kernel.org/all/20211029072424.9109-1-bhe at redhat.com/T/#u
v1->v2:
No code change. The v1 post has been acked by Dave Young but not
merged. In v2, just merging the old patch 1 and 2 as per Dave's concern in v1 post.Baoquan He (2):
x86/kexec: fix memory leak of elf header buffer
kexec_file: clean uparch_kexec_kernel_image_loadarch/x86/kernel/machine_kexec_64.c | 23 +++++++++-------------- include/linux/kexec.h | 1 - kernel/kexec_file.c | 9 ++------- 3 files changed, 11 insertions(+), 22 deletions(-)Signature · 1 line
2.31.1
Andrew Morton
maintainerQuoted 10 lines · On Thu, 31 Mar 2022 19:32:17 +0800 Baoquan He <[email protected]> wrote:
Hi Andrew,
Quoted 4 lines · Baoquan He
The memory leak is reported by kmemleak detector, has been existing for very long time. It casue much memory loss on large machine with huge memory hotplug which will trigger kdump kernel reloading many times, with
kexec_file_loadinterface.Could you merge these two patches? Or should I ping x86 maintainers to take them?
Ah, sorry, I tend to fall asleep if there's "x86" in the subject. Poking a sleeping Andrew is always the right thing to do. Shall look at them.
Quoted 10 lines · Andrew Morton
Quoted 8 lines · On Thu, 31 Mar 2022 19:32:17 +0800 Baoquan He <[email protected]> wrote:
Hi Andrew,
Quoted 4 lines · Baoquan He
The memory leak is reported by kmemleak detector, has been existing for very long time. It casue much memory loss on large machine with huge memory hotplug which will trigger kdump kernel reloading many times, with
kexec_file_loadinterface.
Quoted 6 lines
Quoted 2 lines
Could you merge these two patches? Or should I ping x86 maintainers to take them?
Ah, sorry, I tend to fall asleep if there's "x86" in the subject. Poking a sleeping Andrew is always the right thing to do. Shall look at them.
Thanks, Andrew.
Since it's also MM related, so ping you and x86 maintainers to see who can help pick them.
Borislav Petkov
maintainerQuoted 2 lines · Baoquan He
Since it's also MM related, so ping you and x86 maintainers to see who can help pick them.
I'd leave that decision to akpm. :-)
Thx.
Signature · 4 lines
Regards/Gruss,
Boris.
https://people.kernel.org/tglx/notes-about-netiquette