x86/boot: save fields explicitly, zero out everything else
Subsystems (2)
v2 cover x86/boot: save fields explicitly, zero out everything else
From: John Hubbard <[email protected]>
Hi,
This uses the "save each field explicitly" approach that we discussed during the first review [1]. As in [1], this is motivated by a desire to clear the compiler warnings when building with gcc 9.
This is difficult to properly test. I've done a basic boot test, but if there are actually errors in which items get zeroed or not, I don't have a good test to uncover that.
[1] https://lore.kernel.org/r/[email protected]
John Hubbard (1):
x86/boot: save fields explicitly, zero out everything else
arch/x86/include/asm/bootparam_utils.h | 62 +++++++++++++++++++------- 1 file changed, 47 insertions(+), 15 deletions(-)
Signature · 1 line
2.22.0
PATCH v2 1/1 x86/boot: save fields explicitly, zero out everything else
From: John Hubbard <[email protected]>
Recent gcc compilers (gcc 9.1) generate warnings about an out of bounds memset, if you trying memset across several fields of a struct. This generated a couple of warnings on x86_64 builds.
Fix this by explicitly saving the fields in struct boot_params that are intended to be preserved, and zeroing all the rest.
--- arch/x86/include/asm/bootparam_utils.h | 62 +++++++++++++++++++------- 1 file changed, 47 insertions(+), 15 deletions(-)
arch/x86/include/asm/bootparam_utils.h +47 -15
@@ -18,6 +18,20 @@* Note: efi_info is commonly left uninitialized, but that field has a * private magic, so it is better to leave it unchanged. */ + +#define sizeof_mbr (type, member) ({ sizeof(((type *)0)->member); }) + +#define BOOT_PARAM_PRESERVE (struct_member) \ + { \ + .start = offsetof(struct boot_params, struct_member), \ + .len = sizeof_mbr(struct boot_params, struct_member), \ + } + +struct boot_params_to_save { +unsigned int start; +unsigned int len; +}; +static void sanitize_boot_params (struct boot_params *boot_params) {/*
@@ -35,21 +49,39 @@ static void sanitize_boot_params(struct boot_params *boot_params)* problems again. */ if (boot_params->sentinel) { -/* fields in boot_params are left uninitialized, clear them */ - boot_params->acpi_rsdp_addr =0 ; -memset (&boot_params->ext_ramdisk_image,0 , - (char *)&boot_params->efi_info - - (char *)&boot_params->ext_ramdisk_image); -memset (&boot_params->kbd_status,0 , - (char *)&boot_params->hdr - - (char *)&boot_params->kbd_status); -memset (&boot_params->_pad7[0 ],0 , - (char *)&boot_params->edd_mbr_sig_buffer[0 ] - - (char *)&boot_params->_pad7[0 ]); -memset (&boot_params->_pad8[0 ],0 , - (char *)&boot_params->eddbuf[0 ] - - (char *)&boot_params->_pad8[0 ]); -memset (&boot_params->_pad9[0 ],0 ,sizeof (boot_params->_pad9)); +static struct boot_params scratch; +char *bp_base = (char *)boot_params; +char *save_base = (char *)&scratch; +int i; + +const struct boot_params_to_save to_save[] = { +BOOT_PARAM_PRESERVE (screen_info), +BOOT_PARAM_PRESERVE (apm_bios_info), +BOOT_PARAM_PRESERVE (tboot_addr), +BOOT_PARAM_PRESERVE (ist_info), +BOOT_PARAM_PRESERVE (acpi_rsdp_addr), +BOOT_PARAM_PRESERVE (hd0_info), +BOOT_PARAM_PRESERVE (hd1_info), +BOOT_PARAM_PRESERVE (sys_desc_table), +BOOT_PARAM_PRESERVE (olpc_ofw_header), +BOOT_PARAM_PRESERVE (efi_info), +BOOT_PARAM_PRESERVE (alt_mem_k), +BOOT_PARAM_PRESERVE (scratch), +BOOT_PARAM_PRESERVE (e820_entries), +BOOT_PARAM_PRESERVE (eddbuf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer), +BOOT_PARAM_PRESERVE (e820_table), +BOOT_PARAM_PRESERVE (eddbuf), + }; + +memset (&scratch,0 ,sizeof (scratch)); + +for (i =0 ; i <ARRAY_SIZE (to_save); i++) +memcpy (save_base + to_save[i].start, + bp_base + to_save[i].start, to_save[i].len); + +memcpy (boot_params, save_base,sizeof (*boot_params)); } }
Signature · 1 line
2.22.0
From: [email protected]
Quoted 50 lines
Sent: 31 July 2019 06:46
Recent gcc compilers (gcc 9.1) generate warnings about an out of bounds memset, if you trying memset across several fields of a struct. This generated a couple of warnings on
x86_64builds.Fix this by explicitly saving the fields in struct
boot_paramsthat are intended to be preserved, and zeroing all the rest.Suggested-by Thomas Gleixner <[email protected]>Suggested-by H. Peter Anvin <[email protected]>Signed-off-by John Hubbard <[email protected]>--- arch/x86/include/asm/bootparam_utils.h | 62 +++++++++++++++++++------- 1 file changed, 47 insertions(+), 15 deletions(-)arch/x86/include/asm/bootparam_utils.h +14
-5@@ -18,6 +18,20 @@* Note: efi_info is commonly left uninitialized, but that field has a * private magic, so it is better to leave it unchanged. */ + +#define sizeof_mbr (type, member) ({ sizeof(((type *)0)->member); }) + +#define BOOT_PARAM_PRESERVE (struct_member) \ + { \ + .start = offsetof(struct boot_params, struct_member), \ + .len = sizeof_mbr(struct boot_params, struct_member), \ + } + +struct boot_params_to_save { +unsigned int start; +unsigned int len; +}; +static void sanitize_boot_params (struct boot_params *boot_params) {/* @@ -35,21 +49,39 @@ static void sanitize_boot_params(struct boot_params *boot_params)* problems again. */ if (boot_params->sentinel) { -/* fields in boot_params are left uninitialized, clear them */ - boot_params->acpi_rsdp_addr =0 ; -memset (&boot_params->ext_ramdisk_image,0 , - (char *)&boot_params->efi_info - - (char *)&boot_params->ext_ramdisk_image);
arch/x86/include/asm/bootparam_utils.h static void sanitize_boot_params(struct boot_params *boot_params) · line 52
Quoted 10 lines
arch/x86/include/asm/bootparam_utils.h +0
-10-memset (&boot_params->kbd_status,0 , - (char *)&boot_params->hdr - - (char *)&boot_params->kbd_status); -memset (&boot_params->_pad7[0 ],0 , - (char *)&boot_params->edd_mbr_sig_buffer[0 ] - - (char *)&boot_params->_pad7[0 ]); -memset (&boot_params->_pad8[0 ],0 , - (char *)&boot_params->eddbuf[0 ] - - (char *)&boot_params->_pad8[0 ]); -memset (&boot_params->_pad9[0 ],0 ,sizeof (boot_params->_pad9));
...
How about replacing the above first using:
#define zero_struct_fields(ptr, from, to) memset(&ptr->from, 0, (char *)&ptr->to - (char *)&ptr->from) zero_struct_fields(boot_params, ext_ramdisk_image, efi_info); ...
Which is absolutely identical to the original code.
The replacing the define with:
#define so(s, m) offsetof(typeof(*s), m) #define zero_struct_fields(ptr, from, to) memset((char *)ptr + so(ptr, from), 0, so(ptr, to) - so(ptr, from))
which gcc probably doesn't complain about, but should generate identical code again. There might be an existing define for so().
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)
Quoted 2 lines · David Laight
From: [email protected]
Quoted 1 line
Sent: 31 July 2019 06:46
...
Quoted 22 lines
Quoted 6 lines
arch/x86/include/asm/bootparam_utils.h +0
-5if (boot_params->sentinel) { -/* fields in boot_params are left uninitialized, clear them */ - boot_params->acpi_rsdp_addr =0 ; -memset (&boot_params->ext_ramdisk_image,0 , - (char *)&boot_params->efi_info - - (char *)&boot_params->ext_ramdisk_image);arch/x86/include/asm/bootparam_utils.h static void sanitize_boot_params(struct boot_params *boot_params) · line 52
Quoted 10 lines
arch/x86/include/asm/bootparam_utils.h +0
-10-memset (&boot_params->kbd_status,0 , - (char *)&boot_params->hdr - - (char *)&boot_params->kbd_status); -memset (&boot_params->_pad7[0 ],0 , - (char *)&boot_params->edd_mbr_sig_buffer[0 ] - - (char *)&boot_params->_pad7[0 ]); -memset (&boot_params->_pad8[0 ],0 , - (char *)&boot_params->eddbuf[0 ] - - (char *)&boot_params->_pad8[0 ]); -memset (&boot_params->_pad9[0 ],0 ,sizeof (boot_params->_pad9));...
How about replacing the above first using:
#define zero_struct_fields(ptr, from, to) memset(&ptr->from, 0, (char *)&ptr->to - (char *)&ptr->from) zero_struct_fields(boot_params, ext_ramdisk_image, efi_info); ...
Quoted 8 lines
Which is absolutely identical to the original code.
The replacing the define with:
#define so(s, m) offsetof(typeof(*s), m) #define zero_struct_fields(ptr, from, to) memset((char *)ptr + so(ptr, from), 0, so(ptr, to) - so(ptr, from))which gcc probably doesn't complain about, but should generate identical code again. There might be an existing define for
so().
Hi David,
There was discussion about that [1], but preference ending up being to flip this around, in order to more closely match the original intent of this function (zero out everything except for certain carefully selected fields), and to therefore be more likely to keep working if fields are added.
[1] https://lore.kernel.org/lkml/[email protected]/
thanks,
Signature · 2 lines
John Hubbard NVIDIA
tip-bot for John Hubbard
merge bot[tip:x86/boot] x86/boot: Save fields explicitly, zero out everything else
In reply to [email protected]x86/boot: Save fields explicitly, zero out everything else
Commit-ID: 610666f0581557944c3abec93a7c125b8303442c Gitweb: https://git.kernel.org/tip/610666f0581557944c3abec93a7c125b8303442c Author: John Hubbard <[email protected]> AuthorDate: Tue, 30 Jul 2019 22:46:27 -0700 Committer: Thomas Gleixner <[email protected]> CommitDate: Wed, 7 Aug 2019 15:16:04 +0200
x86/boot: Save fields explicitly, zero out everything else
Recent gcc compilers (gcc 9.1) generate warnings about an out of bounds memset, if you trying memset across several fields of a struct. This generated a couple of warnings on x86_64 builds.
Fix this by explicitly saving the fields in struct boot_params that are intended to be preserved, and zeroing all the rest.
--- arch/x86/include/asm/bootparam_utils.h | 63 ++++++++++++++++++++++++++-------- 1 file changed, 48 insertions(+), 15 deletions(-)
arch/x86/include/asm/bootparam_utils.h +48 -15
@@ -18,6 +18,20 @@* Note: efi_info is commonly left uninitialized, but that field has a * private magic, so it is better to leave it unchanged. */ + +#define sizeof_mbr (type, member) ({ sizeof(((type *)0)->member); }) + +#define BOOT_PARAM_PRESERVE (struct_member) \ + { \ + .start = offsetof(struct boot_params, struct_member), \ + .len = sizeof_mbr(struct boot_params, struct_member), \ + } + +struct boot_params_to_save { +unsigned int start; +unsigned int len; +}; +static void sanitize_boot_params (struct boot_params *boot_params) {/*
@@ -35,21 +49,40 @@ static void sanitize_boot_params(struct boot_params *boot_params)* problems again. */ if (boot_params->sentinel) { -/* fields in boot_params are left uninitialized, clear them */ - boot_params->acpi_rsdp_addr =0 ; -memset (&boot_params->ext_ramdisk_image,0 , - (char *)&boot_params->efi_info - - (char *)&boot_params->ext_ramdisk_image); -memset (&boot_params->kbd_status,0 , - (char *)&boot_params->hdr - - (char *)&boot_params->kbd_status); -memset (&boot_params->_pad7[0 ],0 , - (char *)&boot_params->edd_mbr_sig_buffer[0 ] - - (char *)&boot_params->_pad7[0 ]); -memset (&boot_params->_pad8[0 ],0 , - (char *)&boot_params->eddbuf[0 ] - - (char *)&boot_params->_pad8[0 ]); -memset (&boot_params->_pad9[0 ],0 ,sizeof (boot_params->_pad9)); +static struct boot_params scratch; +char *bp_base = (char *)boot_params; +char *save_base = (char *)&scratch; +int i; + +const struct boot_params_to_save to_save[] = { +BOOT_PARAM_PRESERVE (screen_info), +BOOT_PARAM_PRESERVE (apm_bios_info), +BOOT_PARAM_PRESERVE (tboot_addr), +BOOT_PARAM_PRESERVE (ist_info), +BOOT_PARAM_PRESERVE (acpi_rsdp_addr), +BOOT_PARAM_PRESERVE (hd0_info), +BOOT_PARAM_PRESERVE (hd1_info), +BOOT_PARAM_PRESERVE (sys_desc_table), +BOOT_PARAM_PRESERVE (olpc_ofw_header), +BOOT_PARAM_PRESERVE (efi_info), +BOOT_PARAM_PRESERVE (alt_mem_k), +BOOT_PARAM_PRESERVE (scratch), +BOOT_PARAM_PRESERVE (e820_entries), +BOOT_PARAM_PRESERVE (eddbuf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer), +BOOT_PARAM_PRESERVE (e820_table), +BOOT_PARAM_PRESERVE (eddbuf), + }; + +memset (&scratch,0 ,sizeof (scratch)); + +for (i =0 ; i <ARRAY_SIZE (to_save); i++) { +memcpy (save_base + to_save[i].start, + bp_base + to_save[i].start, to_save[i].len); + } + +memcpy (boot_params, save_base,sizeof (*boot_params)); } }
tip-bot for John Hubbard
merge bot[tip:x86/boot] x86/boot: Save fields explicitly, zero out everything else
In reply to [email protected]x86/boot: Save fields explicitly, zero out everything else
Commit-ID: a156cadef2fe445ac423670eace517b39a01ccd0 Gitweb: https://git.kernel.org/tip/a156cadef2fe445ac423670eace517b39a01ccd0 Author: John Hubbard <[email protected]> AuthorDate: Tue, 30 Jul 2019 22:46:27 -0700 Committer: Thomas Gleixner <[email protected]> CommitDate: Wed, 7 Aug 2019 15:22:53 +0200
x86/boot: Save fields explicitly, zero out everything else
Recent gcc compilers (gcc 9.1) generate warnings about an out of bounds memset, if the memset goes accross several fields of a struct. This generated a couple of warnings on x86_64 builds in sanitize_boot_params().
Fix this by explicitly saving the fields in struct boot_params that are intended to be preserved, and zeroing all the rest.
--- arch/x86/include/asm/bootparam_utils.h | 63 ++++++++++++++++++++++++++-------- 1 file changed, 48 insertions(+), 15 deletions(-)
arch/x86/include/asm/bootparam_utils.h +48 -15
@@ -18,6 +18,20 @@* Note: efi_info is commonly left uninitialized, but that field has a * private magic, so it is better to leave it unchanged. */ + +#define sizeof_mbr (type, member) ({ sizeof(((type *)0)->member); }) + +#define BOOT_PARAM_PRESERVE (struct_member) \ + { \ + .start = offsetof(struct boot_params, struct_member), \ + .len = sizeof_mbr(struct boot_params, struct_member), \ + } + +struct boot_params_to_save { +unsigned int start; +unsigned int len; +}; +static void sanitize_boot_params (struct boot_params *boot_params) {/*
@@ -35,21 +49,40 @@ static void sanitize_boot_params(struct boot_params *boot_params)* problems again. */ if (boot_params->sentinel) { -/* fields in boot_params are left uninitialized, clear them */ - boot_params->acpi_rsdp_addr =0 ; -memset (&boot_params->ext_ramdisk_image,0 , - (char *)&boot_params->efi_info - - (char *)&boot_params->ext_ramdisk_image); -memset (&boot_params->kbd_status,0 , - (char *)&boot_params->hdr - - (char *)&boot_params->kbd_status); -memset (&boot_params->_pad7[0 ],0 , - (char *)&boot_params->edd_mbr_sig_buffer[0 ] - - (char *)&boot_params->_pad7[0 ]); -memset (&boot_params->_pad8[0 ],0 , - (char *)&boot_params->eddbuf[0 ] - - (char *)&boot_params->_pad8[0 ]); -memset (&boot_params->_pad9[0 ],0 ,sizeof (boot_params->_pad9)); +static struct boot_params scratch; +char *bp_base = (char *)boot_params; +char *save_base = (char *)&scratch; +int i; + +const struct boot_params_to_save to_save[] = { +BOOT_PARAM_PRESERVE (screen_info), +BOOT_PARAM_PRESERVE (apm_bios_info), +BOOT_PARAM_PRESERVE (tboot_addr), +BOOT_PARAM_PRESERVE (ist_info), +BOOT_PARAM_PRESERVE (acpi_rsdp_addr), +BOOT_PARAM_PRESERVE (hd0_info), +BOOT_PARAM_PRESERVE (hd1_info), +BOOT_PARAM_PRESERVE (sys_desc_table), +BOOT_PARAM_PRESERVE (olpc_ofw_header), +BOOT_PARAM_PRESERVE (efi_info), +BOOT_PARAM_PRESERVE (alt_mem_k), +BOOT_PARAM_PRESERVE (scratch), +BOOT_PARAM_PRESERVE (e820_entries), +BOOT_PARAM_PRESERVE (eddbuf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer), +BOOT_PARAM_PRESERVE (e820_table), +BOOT_PARAM_PRESERVE (eddbuf), + }; + +memset (&scratch,0 ,sizeof (scratch)); + +for (i =0 ; i <ARRAY_SIZE (to_save); i++) { +memcpy (save_base + to_save[i].start, + bp_base + to_save[i].start, to_save[i].len); + } + +memcpy (boot_params, save_base,sizeof (*boot_params)); } }
Quoted 9 lines · On 31/07/2019 06:46, [email protected] wrote:
From: John Hubbard <[email protected]>
Recent gcc compilers (gcc 9.1) generate warnings about an out of bounds memset, if you trying memset across several fields of a struct. This generated a couple of warnings on
x86_64builds.Fix this by explicitly saving the fields in struct
boot_paramsthat are intended to be preserved, and zeroing all the rest.
I applied John's patch below to v5.3-rc3-285-gecb095bff5d4 and have been running the resultant kernel for two days now, including 7 or 8 cold starts and reboots. The warnings that were produced by gcc9 are no longer emitted and, other than a pre-existing problem (no network after resume from suspend or hibernate which I will investigate and, if necessary, report later today), the kernel has supported my typical day to day activities (building software, email, browsing, listening to music, watching video) without problem.
Quoted 88 lines
Suggested-by Thomas Gleixner <[email protected]>Suggested-by H. Peter Anvin <[email protected]>Signed-off-by John Hubbard <[email protected]>--- arch/x86/include/asm/bootparam_utils.h | 62 +++++++++++++++++++------- 1 file changed, 47 insertions(+), 15 deletions(-)arch/x86/include/asm/bootparam_utils.h +47
-15@@ -18,6 +18,20 @@* Note: efi_info is commonly left uninitialized, but that field has a * private magic, so it is better to leave it unchanged. */ + +#define sizeof_mbr (type, member) ({ sizeof(((type *)0)->member); }) + +#define BOOT_PARAM_PRESERVE (struct_member) \ + { \ + .start = offsetof(struct boot_params, struct_member), \ + .len = sizeof_mbr(struct boot_params, struct_member), \ + } + +struct boot_params_to_save { +unsigned int start; +unsigned int len; +}; +static void sanitize_boot_params (struct boot_params *boot_params) {/* @@ -35,21 +49,39 @@ static void sanitize_boot_params(struct boot_params *boot_params)* problems again. */ if (boot_params->sentinel) { -/* fields in boot_params are left uninitialized, clear them */ - boot_params->acpi_rsdp_addr =0 ; -memset (&boot_params->ext_ramdisk_image,0 , - (char *)&boot_params->efi_info - - (char *)&boot_params->ext_ramdisk_image); -memset (&boot_params->kbd_status,0 , - (char *)&boot_params->hdr - - (char *)&boot_params->kbd_status); -memset (&boot_params->_pad7[0 ],0 , - (char *)&boot_params->edd_mbr_sig_buffer[0 ] - - (char *)&boot_params->_pad7[0 ]); -memset (&boot_params->_pad8[0 ],0 , - (char *)&boot_params->eddbuf[0 ] - - (char *)&boot_params->_pad8[0 ]); -memset (&boot_params->_pad9[0 ],0 ,sizeof (boot_params->_pad9)); +static struct boot_params scratch; +char *bp_base = (char *)boot_params; +char *save_base = (char *)&scratch; +int i; + +const struct boot_params_to_save to_save[] = { +BOOT_PARAM_PRESERVE (screen_info), +BOOT_PARAM_PRESERVE (apm_bios_info), +BOOT_PARAM_PRESERVE (tboot_addr), +BOOT_PARAM_PRESERVE (ist_info), +BOOT_PARAM_PRESERVE (acpi_rsdp_addr), +BOOT_PARAM_PRESERVE (hd0_info), +BOOT_PARAM_PRESERVE (hd1_info), +BOOT_PARAM_PRESERVE (sys_desc_table), +BOOT_PARAM_PRESERVE (olpc_ofw_header), +BOOT_PARAM_PRESERVE (efi_info), +BOOT_PARAM_PRESERVE (alt_mem_k), +BOOT_PARAM_PRESERVE (scratch), +BOOT_PARAM_PRESERVE (e820_entries), +BOOT_PARAM_PRESERVE (eddbuf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer), +BOOT_PARAM_PRESERVE (e820_table), +BOOT_PARAM_PRESERVE (eddbuf), + }; + +memset (&scratch,0 ,sizeof (scratch)); + +for (i =0 ; i <ARRAY_SIZE (to_save); i++) +memcpy (save_base + to_save[i].start, + bp_base + to_save[i].start, to_save[i].len); + +memcpy (boot_params, save_base,sizeof (*boot_params)); } }
tip-bot for John Hubbard
merge bot[tip:x86/urgent] x86/boot: Save fields explicitly, zero out everything else
In reply to [email protected]x86/boot: Save fields explicitly, zero out everything else
Commit-ID: a90118c445cc7f07781de26a9684d4ec58bfcfd1 Gitweb: https://git.kernel.org/tip/a90118c445cc7f07781de26a9684d4ec58bfcfd1 Author: John Hubbard <[email protected]> AuthorDate: Tue, 30 Jul 2019 22:46:27 -0700 Committer: Thomas Gleixner <[email protected]> CommitDate: Fri, 16 Aug 2019 14:20:00 +0200
x86/boot: Save fields explicitly, zero out everything else
Recent gcc compilers (gcc 9.1) generate warnings about an out of bounds memset, if the memset goes accross several fields of a struct. This generated a couple of warnings on x86_64 builds in sanitize_boot_params().
Fix this by explicitly saving the fields in struct boot_params that are intended to be preserved, and zeroing all the rest.
[ tglx: Tagged for stable as it breaks the warning free build there as well ]
--- arch/x86/include/asm/bootparam_utils.h | 63 ++++++++++++++++++++++++++-------- 1 file changed, 48 insertions(+), 15 deletions(-)
arch/x86/include/asm/bootparam_utils.h +48 -15
@@ -18,6 +18,20 @@* Note: efi_info is commonly left uninitialized, but that field has a * private magic, so it is better to leave it unchanged. */ + +#define sizeof_mbr (type, member) ({ sizeof(((type *)0)->member); }) + +#define BOOT_PARAM_PRESERVE (struct_member) \ + { \ + .start = offsetof(struct boot_params, struct_member), \ + .len = sizeof_mbr(struct boot_params, struct_member), \ + } + +struct boot_params_to_save { +unsigned int start; +unsigned int len; +}; +static void sanitize_boot_params (struct boot_params *boot_params) {/*
@@ -35,21 +49,40 @@ static void sanitize_boot_params(struct boot_params *boot_params)* problems again. */ if (boot_params->sentinel) { -/* fields in boot_params are left uninitialized, clear them */ - boot_params->acpi_rsdp_addr =0 ; -memset (&boot_params->ext_ramdisk_image,0 , - (char *)&boot_params->efi_info - - (char *)&boot_params->ext_ramdisk_image); -memset (&boot_params->kbd_status,0 , - (char *)&boot_params->hdr - - (char *)&boot_params->kbd_status); -memset (&boot_params->_pad7[0 ],0 , - (char *)&boot_params->edd_mbr_sig_buffer[0 ] - - (char *)&boot_params->_pad7[0 ]); -memset (&boot_params->_pad8[0 ],0 , - (char *)&boot_params->eddbuf[0 ] - - (char *)&boot_params->_pad8[0 ]); -memset (&boot_params->_pad9[0 ],0 ,sizeof (boot_params->_pad9)); +static struct boot_params scratch; +char *bp_base = (char *)boot_params; +char *save_base = (char *)&scratch; +int i; + +const struct boot_params_to_save to_save[] = { +BOOT_PARAM_PRESERVE (screen_info), +BOOT_PARAM_PRESERVE (apm_bios_info), +BOOT_PARAM_PRESERVE (tboot_addr), +BOOT_PARAM_PRESERVE (ist_info), +BOOT_PARAM_PRESERVE (acpi_rsdp_addr), +BOOT_PARAM_PRESERVE (hd0_info), +BOOT_PARAM_PRESERVE (hd1_info), +BOOT_PARAM_PRESERVE (sys_desc_table), +BOOT_PARAM_PRESERVE (olpc_ofw_header), +BOOT_PARAM_PRESERVE (efi_info), +BOOT_PARAM_PRESERVE (alt_mem_k), +BOOT_PARAM_PRESERVE (scratch), +BOOT_PARAM_PRESERVE (e820_entries), +BOOT_PARAM_PRESERVE (eddbuf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries), +BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer), +BOOT_PARAM_PRESERVE (e820_table), +BOOT_PARAM_PRESERVE (eddbuf), + }; + +memset (&scratch,0 ,sizeof (scratch)); + +for (i =0 ; i <ARRAY_SIZE (to_save); i++) { +memcpy (save_base + to_save[i].start, + bp_base + to_save[i].start, to_save[i].len); + } + +memcpy (boot_params, save_base,sizeof (*boot_params)); } }
PATCH 1/1 x86/boot: Fix regression--secure boot info loss from bootparam sanitizing
In reply to [email protected]From: "John S. Gruber" <[email protected]>
commit a90118c445cc ("x86/boot: Save fields explicitly, zero out everything else") now zeros the secure boot information passed by the boot loader or by the kernel's efi handover mechanism.
Include boot-params.secure_boot in the preserve field list.
---
I noted a change in my computers between running signed 5.3-rc4 and 5.3-rc6 with signed kernels using the efi handoff protocol with grub. The kernel log message "Secure boot enabled" becomes "Secure boot could not be determined". The efi_main function in arch/x86/boot/compressed/eboot.c sets this field early but it is subsequently zeroed by the above referenced commit in the file arch/x86/include/asm/bootparam_utils.h
Applies to 5.3-rc6.
arch/x86/include/asm/bootparam_utils.h | 1 + 1 file changed, 1 insertion(+)
(quoted diff) +0 -0
b/arch/x86/include/asm/bootparam_utils.h
index 9e5f3c7..981fe92 100644
arch/x86/include/asm/bootparam_utils.h +0 -0
@@ -70,6 +70,7 @@ static void sanitize_boot_params(struct boot_params
*boot_params)
BOOT_PARAM_PRESERVE(eddbuf_entries), BOOT_PARAM_PRESERVE(edd_mbr_sig_buf_entries), BOOT_PARAM_PRESERVE(edd_mbr_sig_buffer),
+ BOOT_PARAM_PRESERVE(secure_boot),
BOOT_PARAM_PRESERVE(hdr), BOOT_PARAM_PRESERVE(e820_table), BOOT_PARAM_PRESERVE(eddbuf),
Signature · 1 line
2.7.4
Re: [PATCH] x86/boot: Fix regression--secure boot info loss from bootparam sanitizing
Quoted 11 lines · John S Gruber
From: "John S. Gruber" <[email protected]>
commit
a90118c445cc("x86/boot: Save fields explicitly, zero out everything else") now zeros the secure boot information passed by the boot loader or by the kernel's efi handover mechanism.Include boot-params.secure_boot in the preserve field list.
Signed-off-by John S. Gruber <[email protected]>---
Quoted 9 lines
I noted a change in my computers between running signed 5.3-rc4 and 5.3-rc6 with signed kernels using the efi handoff protocol with grub. The kernel log message "Secure boot enabled" becomes "Secure boot could not be determined". The
efi_mainfunction inarch/x86/boot/compressed/eboot.csets this field early but it is subsequently zeroed by the above referenced commit in the filearch/x86/include/asm/bootparam_utils.hApplies to 5.3-rc6.
Hi,
The fix itself looks good, so you can add:
...but note that the commit description should get a few tweaks:
1. Your description above is actually well-suited for the commit log, so please add that in. Especially the symptoms are desirable to have on record.
2. This should Cc: [email protected], because the whole thing made it into -stable and those kernels need this fix.
3. Also need a Fixes tag:
thanks,
Signature · 2 lines
John Hubbard NVIDIA
Quoted 18 lines
arch/x86/include/asm/bootparam_utils.h | 1 + 1 file changed, 1 insertion(+)arch/x86/include/asm/bootparam_utils.h +1
-0@@ -70,6 +70,7 @@ static void sanitize_boot_params(struct boot_params *boot_params)BOOT_PARAM_PRESERVE (eddbuf_entries),BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries),BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer), +BOOT_PARAM_PRESERVE (secure_boot),BOOT_PARAM_PRESERVE (hdr), BOOT_PARAM_PRESERVE(e820_table),
BOOT_PARAM_PRESERVE(eddbuf),
PATCH v2 1/1 x86/boot: Fix regression--secure boot info loss from bootparam sanitizing
In reply to [email protected]From: "John S. Gruber" <[email protected]>
commit a90118c445cc ("x86/boot: Save fields explicitly, zero out everything else") now zeros the secure boot information passed by the boot loader or by the kernel's efi handover mechanism. Include boot-params.secure_boot in the preserve field list.
I noted a change in my computers between running signed 5.3-rc4 and 5.3-rc6 with signed kernels using the efi handoff protocol with grub. The kernel log message "Secure boot enabled" becomes "Secure boot could not be determined". The efi_main function in arch/x86/boot/compressed/eboot.c sets this field early but it is subsequently zeroed by the above referenced commit in the file arch/x86/include/asm/bootparam_utils.h
out everything else")
---
Adjusted the patch for John Hubbard's comments.
arch/x86/include/asm/bootparam_utils.h | 1 + 1 file changed, 1 insertion(+)
(quoted diff) +0 -0
b/arch/x86/include/asm/bootparam_utils.h
index 9e5f3c7..981fe92 100644
arch/x86/include/asm/bootparam_utils.h +0 -0
@@ -70,6 +70,7 @@ static void sanitize_boot_params(struct boot_params
*boot_params)
BOOT_PARAM_PRESERVE(eddbuf_entries), BOOT_PARAM_PRESERVE(edd_mbr_sig_buf_entries), BOOT_PARAM_PRESERVE(edd_mbr_sig_buffer),
+ BOOT_PARAM_PRESERVE(secure_boot),
BOOT_PARAM_PRESERVE(hdr), BOOT_PARAM_PRESERVE(e820_table), BOOT_PARAM_PRESERVE(eddbuf),
Signature · 1 line
2.7.4
Borislav Petkov
maintainerRe: [PATCH V2] x86/boot: Fix regression--secure boot info loss from bootparam sanitizing
Quoted 30 lines · John S Gruber
From: "John S. Gruber" <[email protected]>
commit
a90118c445cc("x86/boot: Save fields explicitly, zero out everything else") now zeros the secure boot information passed by the boot loader or by the kernel's efi handover mechanism. Include boot-params.secure_boot in the preserve field list.I noted a change in my computers between running signed 5.3-rc4 and 5.3-rc6 with signed kernels using the efi handoff protocol with grub. The kernel log message "Secure boot enabled" becomes "Secure boot could not be determined". The
efi_mainfunction inarch/x86/boot/compressed/eboot.csets this field early but it is subsequently zeroed by the above referenced commit in the filearch/x86/include/asm/bootparam_utils.hFixes commit a90118c445cc ("x86/boot: Save fields explicitly, zeroout everything else")
Signed-off-by John S. Gruber <[email protected]>---
Adjusted the patch for John Hubbard's comments.
arch/x86/include/asm/bootparam_utils.h | 1 + 1 file changed, 1 insertion(+)arch/x86/include/asm/bootparam_utils.h +0
-0
arch/x86/include/asm/bootparam_utils.h static void sanitize_boot_params(struct boot_params · line 70
Quoted 1 line
arch/x86/include/asm/bootparam_utils.h +0
-0@@ -70,6 +70,7 @@ static void sanitize_boot_params(struct boot_params *boot_params)
gmail has managed to chew this patch:
checking file arch/x86/include/asm/bootparam_utils.h
patch: **** malformed patch at line 48: *boot_params)
See: https://www.kernel.org/doc/html/latest/process/email-clients.html#gmail-web-gui
You might find a better client in there if you wanna send more patches in the future.
Signature · 4 lines
Regards/Gruss,
Boris.
Good mailing practices for 400: avoid top-posting and trim the reply.[tip: x86/urgent] x86/boot: Preserve boot_params.secure_boot from sanitizing
In reply to John S GruberThe following commit has been merged into the x86/urgent branch of tip:
The following commit has been merged into the x86/urgent branch of tip:
Commit-ID: 29d9a0b50736768f042752070e5cdf4e4d4c00df Gitweb: https://git.kernel.org/tip/29d9a0b50736768f042752070e5cdf4e4d4c00df Author: John S. Gruber <[email protected]> AuthorDate: Mon, 02 Sep 2019 00:00:54 +02:00 Committer: Borislav Petkov <[email protected]> CommitterDate: Mon, 02 Sep 2019 09:17:45 +02:00
x86/boot: Preserve boot_params.secure_boot from sanitizing
Commit
a90118c445cc ("x86/boot: Save fields explicitly, zero out everything else")
now zeroes the secure boot setting information (enabled/disabled/...) passed by the boot loader or by the kernel's EFI handover mechanism.
The problem manifests itself with signed kernels using the EFI handoff protocol with grub and the kernel loses the information whether secure boot is enabled in the firmware, i.e., the log message "Secure boot enabled" becomes "Secure boot could not be determined".
efi_main() arch/x86/boot/compressed/eboot.c sets this field early but it is subsequently zeroed by the above referenced commit.
Include boot_params.secure_boot in the preserve field list.
[ bp: restructure commit message and massage. ]
--- arch/x86/include/asm/bootparam_utils.h | 1 + 1 file changed, 1 insertion(+)
arch/x86/include/asm/bootparam_utils.h +1 -0
@@ -70,6 +70,7 @@ static void sanitize_boot_params(struct boot_params *boot_params)BOOT_PARAM_PRESERVE (eddbuf_entries),BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries),BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer), +BOOT_PARAM_PRESERVE (secure_boot),BOOT_PARAM_PRESERVE (hdr),BOOT_PARAM_PRESERVE (e820_table), BOOT_PARAM_PRESERVE(eddbuf),
PATCH 1/1 x86/boot: v4.4 stable and v4.9 stable boot failure due to dropped patch line
In reply to [email protected]This regards upstream commit a90118c445cc ("x86/boot: Save fields explicitly, zero out everything else") application to linux-stable.
Its corresponding commits to the stable 4.4 and 4.9 trees didn't apply correctly, probably due to a field name change (e820_table had been named e820_map before 4.10).
On my desktop I'm unable to boot a signed kernel due to these commits.
Add e820_map (to replace e820_table) to the preserved fields so that the E820 memory regions in boot_params can be accessed by the kernel after boot_params has been sanitized.
---
I tested stable 4.14.145, 4.19.74, and 5.2.16 successfully under the same circumstances. Only 4.4 and 4.9 are affected by this dropped line.
arch/x86/include/asm/bootparam_utils.h | 1 + 1 file changed, 1 insertion(+)
arch/x86/include/asm/bootparam_utils.h +1 -0
@@ -71,6 +71,7 @@ static void sanitize_boot_params(struct boot_params *boot_params)BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries),BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer),BOOT_PARAM_PRESERVE (hdr), +BOOT_PARAM_PRESERVE (e820_map), BOOT_PARAM_PRESERVE(eddbuf), };
Signature · 1 line
2.7.4
Re: [PATCH] x86/boot: v4.4 stable and v4.9 stable boot failure due to dropped patch line
Quoted 31 lines · John S Gruber
This regards upstream commit
a90118c445cc("x86/boot: Save fields explicitly, zero out everything else") application to linux-stable.Its corresponding commits to the stable 4.4 and 4.9 trees didn't apply correctly, probably due to a field name change (
e820_tablehad been namede820_mapbefore 4.10).On my desktop I'm unable to boot a signed kernel due to these commits.
Add
e820_map(to replacee820_table) to the preserved fields so that the E820 memory regions inboot_paramscan be accessed by the kernel afterboot_paramshas been sanitized.Signed-off-by John S Gruber <[email protected]>Fixes 41664b97f46e ("x86/boot: Save fields explicitly, zero out everything else")Fixes 4e478cb2ccdd ("x86/boot: Save fields explicitly, zero out everything else")---
I tested stable 4.14.145, 4.19.74, and 5.2.16 successfully under the same circumstances. Only 4.4 and 4.9 are affected by this dropped line.
arch/x86/include/asm/bootparam_utils.h | 1 + 1 file changed, 1 insertion(+)arch/x86/include/asm/bootparam_utils.h +0
-0
arch/x86/include/asm/bootparam_utils.h static void sanitize_boot_params(struct boot_params *boot_params) · line 77
Quoted 7 lines
arch/x86/include/asm/bootparam_utils.h +1
-0@@ -71,6 +71,7 @@ static void sanitize_boot_params(struct boot_params *boot_params)BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries),BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer),BOOT_PARAM_PRESERVE (hdr), +BOOT_PARAM_PRESERVE (e820_map), BOOT_PARAM_PRESERVE(eddbuf), };
Thanks for getting -stable figured out and fixed--this has not been smooth sailing!
thanks,
Signature · 2 lines
John Hubbard NVIDIA
Greg Kroah-Hartman
maintainerRe: [PATCH] x86/boot: v4.4 stable and v4.9 stable boot failure due to dropped patch line
Quoted 42 lines · John Hubbard
Quoted 31 lines · John S Gruber
This regards upstream commit
a90118c445cc("x86/boot: Save fields explicitly, zero out everything else") application to linux-stable.Its corresponding commits to the stable 4.4 and 4.9 trees didn't apply correctly, probably due to a field name change (
e820_tablehad been namede820_mapbefore 4.10).On my desktop I'm unable to boot a signed kernel due to these commits.
Add
e820_map(to replacee820_table) to the preserved fields so that the E820 memory regions inboot_paramscan be accessed by the kernel afterboot_paramshas been sanitized.Signed-off-by John S Gruber <[email protected]>Fixes 41664b97f46e ("x86/boot: Save fields explicitly, zero out everything else")Fixes 4e478cb2ccdd ("x86/boot: Save fields explicitly, zero out everything else")---
I tested stable 4.14.145, 4.19.74, and 5.2.16 successfully under the same circumstances. Only 4.4 and 4.9 are affected by this dropped line.
arch/x86/include/asm/bootparam_utils.h | 1 + 1 file changed, 1 insertion(+)arch/x86/include/asm/bootparam_utils.h +0
-0arch/x86/include/asm/bootparam_utils.h static void sanitize_boot_params(struct boot_params *boot_params) · line 77
Quoted 7 lines
arch/x86/include/asm/bootparam_utils.h +1
-0@@ -71,6 +71,7 @@ static void sanitize_boot_params(struct boot_params *boot_params)BOOT_PARAM_PRESERVE (edd_mbr_sig_buf_entries),BOOT_PARAM_PRESERVE (edd_mbr_sig_buffer),BOOT_PARAM_PRESERVE (hdr), +BOOT_PARAM_PRESERVE (e820_map), BOOT_PARAM_PRESERVE(eddbuf), };Reviewed-by John Hubbard <[email protected]>
Quoted 1 line
Thanks for getting -stable figured out and fixed--this has not been smooth sailing!
This change is already in the 4.4.y-rc and 4.9.y-rc releases that went out two days ago for review. So this should be fixed in the next release. If not, please let me know.
thanks,
greg k-h
Quoted 11 lines · On 7/30/19 10:46 PM, [email protected] wrote:
From: John Hubbard <[email protected]>
Hi,
This uses the "save each field explicitly" approach that we discussed during the first review [1]. As in [1], this is motivated by a desire to clear the compiler warnings when building with gcc 9.
This is difficult to properly test. I've done a basic boot test, but if there are actually errors in which items get zeroed or not, I don't have a good test to uncover that.
Also, if anyone has advice about any extra testing I could run on this, please send it my way.
thanks,
Signature · 2 lines
John Hubbard NVIDIA
Quoted 9 lines
[1] https://lore.kernel.org/r/[email protected]
John Hubbard (1):
x86/boot: save fields explicitly, zero out everything elsearch/x86/include/asm/bootparam_utils.h | 62 +++++++++++++++++++------- 1 file changed, 47 insertions(+), 15 deletions(-)