Skip to contentSkip to conversation

ARM, xtensa: highmem: avoid clobbering non-page aligned memory reservations

PATCH In review Mike Rapoport 4 participants 12 messages Oct 31, 2020 – Nov 4, 2020

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ARM, xtensa: highmem: avoid clobbering non-page aligned memory reservations

From: Ard Biesheuvel <[email protected]>

free_highpages() iterates over the free memblock regions in high memory, and marks each page as available for the memory management system.

Until commit cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages") it rounded beginning of each region upwards and end of each region downwards.

However, after that commit free_highmem() rounds the beginning and end of each region downwards, and we may end up freeing a page that is memblock_reserve()d, resulting in memory corruption.

Restore the original rounding of the region boundaries to avoid freeing reserved pages.

Fixes cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages")

---

Max, Russell,

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

v2: fix words order in the commit message

 arch/arm/mm/init.c    | 4 ++--
 arch/xtensa/mm/init.c | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
arch/arm/mm/init.c +2 -2
index d57112a276f5..c23dbf8bebee 100644
@@ -354,8 +354,8 @@ static void __init free_highpages(void)
 	/* set highmem page free */
 	for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE,
 				&range_start, &range_end, NULL) {
-		unsigned long start = PHYS_PFN(range_start);
-		unsigned long end = PHYS_PFN(range_end);
+		unsigned long start = PFN_UP(range_start);
+		unsigned long end = PFN_DOWN(range_end);
 
 		/* Ignore complete lowmem entries */
 		if (end <= max_low)
arch/xtensa/mm/init.c +2 -2
index c6fc83efee0c..8731b7ad9308 100644
@@ -89,8 +89,8 @@ static void __init free_highpages(void)
 	/* set highmem page free */
 	for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE,
 				&range_start, &range_end, NULL) {
-		unsigned long start = PHYS_PFN(range_start);
-		unsigned long end = PHYS_PFN(range_end);
+		unsigned long start = PFN_UP(range_start);
+		unsigned long end = PFN_DOWN(range_end);
 
 		/* Ignore complete lowmem entries */
 		if (end <= max_low)
Signature · 1 line
2.28.0
#1 lore raw
Quoted 22 lines · Mike Rapoport

From: Ard Biesheuvel <[email protected]>

free_highpages() iterates over the free memblock regions in high memory, and marks each page as available for the memory management system.

Until commit cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages") it rounded beginning of each region upwards and end of each region downwards.

However, after that commit free_highmem() rounds the beginning and end of each region downwards, and we may end up freeing a page that is memblock_reserve()d, resulting in memory corruption.

Restore the original rounding of the region boundaries to avoid freeing reserved pages.

Fixes cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages")
Quoted 8 lines

---

Max, Russell,

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

v2: fix words order in the commit message

I really don't understand what is going on here; there seems to be a total disconnect of communication between yourself and Ard. Ard has already submitted a different patch for this to the patch system already, sent yesterday.

https://www.armlinux.org.uk/developer/patches/viewpatch.php?id=9021/1

Please discuss between yourselves how you want to solve the problem, and then submit an agreed and tested patch to those of us upstream; please don't make it for those upstream to pick one of your patches as you are at present.

Thanks.

Quoted 39 lines
 arch/arm/mm/init.c    | 4 ++--
 arch/xtensa/mm/init.c | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
arch/arm/mm/init.c +2 -2
index d57112a276f5..c23dbf8bebee 100644
@@ -354,8 +354,8 @@ static void __init free_highpages(void)
 	/* set highmem page free */
 	for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE,
 				&range_start, &range_end, NULL) {
-		unsigned long start = PHYS_PFN(range_start);
-		unsigned long end = PHYS_PFN(range_end);
+		unsigned long start = PFN_UP(range_start);
+		unsigned long end = PFN_DOWN(range_end);
 
 		/* Ignore complete lowmem entries */
 		if (end <= max_low)
arch/xtensa/mm/init.c +2 -2
index c6fc83efee0c..8731b7ad9308 100644
@@ -89,8 +89,8 @@ static void __init free_highpages(void)
 	/* set highmem page free */
 	for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE,
 				&range_start, &range_end, NULL) {
-		unsigned long start = PHYS_PFN(range_start);
-		unsigned long end = PHYS_PFN(range_end);
+		unsigned long start = PFN_UP(range_start);
+		unsigned long end = PFN_DOWN(range_end);
 
 		/* Ignore complete lowmem entries */
 		if (end <= max_low)
Signature · 1 line
2.28.0
Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
#2 lore raw
Quoted 38 lines · Russell King - ARM Linux admin
Quoted 22 lines · Mike Rapoport

From: Ard Biesheuvel <[email protected]>

free_highpages() iterates over the free memblock regions in high memory, and marks each page as available for the memory management system.

Until commit cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages") it rounded beginning of each region upwards and end of each region downwards.

However, after that commit free_highmem() rounds the beginning and end of each region downwards, and we may end up freeing a page that is memblock_reserve()d, resulting in memory corruption.

Restore the original rounding of the region boundaries to avoid freeing reserved pages.

Fixes cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages")
Quoted 8 lines

---

Max, Russell,

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

v2: fix words order in the commit message

I really don't understand what is going on here; there seems to be a total disconnect of communication between yourself and Ard. Ard has already submitted a different patch for this to the patch system already, sent yesterday.

Quoted 7 lines

https://www.armlinux.org.uk/developer/patches/viewpatch.php?id=9021/1

Please discuss between yourselves how you want to solve the problem, and then submit an agreed and tested patch to those of us upstream; please don't make it for those upstream to pick one of your patches as you are at present.

Apologies for creating this confusion. I posted a patch and dropped it into the patch system when I found the bug.

However, only when Florian asked about a 'fixes' tag, I went back to the history, and realized that the issue was introduced by Mike during the most recent merge window, and affects xtensa as well.

I don't have a preference which patch gets applied, though, so please indicate your preference, and we will adapt accordingly.

Quoted 43 lines
Quoted 36 lines
 arch/arm/mm/init.c    | 4 ++--
 arch/xtensa/mm/init.c | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
arch/arm/mm/init.c +2 -2
index d57112a276f5..c23dbf8bebee 100644
@@ -354,8 +354,8 @@ static void __init free_highpages(void)
      /* set highmem page free */
      for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE,
                              &range_start, &range_end, NULL) {
-             unsigned long start = PHYS_PFN(range_start);
-             unsigned long end = PHYS_PFN(range_end);
+             unsigned long start = PFN_UP(range_start);
+             unsigned long end = PFN_DOWN(range_end);
 
              /* Ignore complete lowmem entries */
              if (end <= max_low)
arch/xtensa/mm/init.c +2 -3
index c6fc83efee0c..8731b7ad9308 100644
@@ -89,8 +89,8 @@ static void __init free_highpages(void)
      /* set highmem page free */
      for_each_free_mem_range(i, NUMA_NO_NODE, MEMBLOCK_NONE,
                              &range_start, &range_end, NULL) {
-             unsigned long start = PHYS_PFN(range_start);
-             unsigned long end = PHYS_PFN(range_end);
+             unsigned long start = PFN_UP(range_start);
+             unsigned long end = PFN_DOWN(range_end);
 
              /* Ignore complete lowmem entries */
              if (end <= max_low)
--
Quoted 3 lines

2.28.0

--
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!

#3 lore raw
Quoted 48 lines · Ard Biesheuvel
Quoted 38 lines · Russell King - ARM Linux admin
Quoted 22 lines · Mike Rapoport

From: Ard Biesheuvel <[email protected]>

free_highpages() iterates over the free memblock regions in high memory, and marks each page as available for the memory management system.

Until commit cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages") it rounded beginning of each region upwards and end of each region downwards.

However, after that commit free_highmem() rounds the beginning and end of each region downwards, and we may end up freeing a page that is memblock_reserve()d, resulting in memory corruption.

Restore the original rounding of the region boundaries to avoid freeing reserved pages.

Fixes cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages")
Quoted 8 lines

---

Max, Russell,

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

v2: fix words order in the commit message

I really don't understand what is going on here; there seems to be a total disconnect of communication between yourself and Ard. Ard has already submitted a different patch for this to the patch system already, sent yesterday.

Quoted 7 lines

https://www.armlinux.org.uk/developer/patches/viewpatch.php?id=9021/1

Please discuss between yourselves how you want to solve the problem, and then submit an agreed and tested patch to those of us upstream; please don't make it for those upstream to pick one of your patches as you are at present.

Quoted 6 lines

Apologies for creating this confusion. I posted a patch and dropped it into the patch system when I found the bug.

However, only when Florian asked about a 'fixes' tag, I went back to the history, and realized that the issue was introduced by Mike during the most recent merge window, and affects xtensa as well.

So why does Mike's patch have:

in it? It seems you haven't been directly involved in Mike's patch.

There's something /really/ not right with the process behind this patch.

Quoted 2 lines

I don't have a preference which patch gets applied, though, so please indicate your preference, and we will adapt accordingly.

I asked for you both to come to a concensus about how you want to proceed, and now you're throwing it back on to me to solve your(pl) mis-communication issue. We haven't heard from Mike yet.

Clearly, I wasn't blunt and stroppy enough to be properly understood. Sort it out between yourselves and tell me which patch you want me to apply.

Signature · 2 lines
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 40Mbps down 10Mbps up. Decent connectivity at last!
#4 lore raw
Quoted 57 lines · Russell King - ARM Linux admin
Quoted 48 lines · Ard Biesheuvel
Quoted 38 lines · Russell King - ARM Linux admin
Quoted 22 lines · Mike Rapoport

From: Ard Biesheuvel <[email protected]>

free_highpages() iterates over the free memblock regions in high memory, and marks each page as available for the memory management system.

Until commit cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages") it rounded beginning of each region upwards and end of each region downwards.

However, after that commit free_highmem() rounds the beginning and end of each region downwards, and we may end up freeing a page that is memblock_reserve()d, resulting in memory corruption.

Restore the original rounding of the region boundaries to avoid freeing reserved pages.

Fixes cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages")
Quoted 8 lines

---

Max, Russell,

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

v2: fix words order in the commit message

I really don't understand what is going on here; there seems to be a total disconnect of communication between yourself and Ard. Ard has already submitted a different patch for this to the patch system already, sent yesterday.

Quoted 7 lines

https://www.armlinux.org.uk/developer/patches/viewpatch.php?id=9021/1

Please discuss between yourselves how you want to solve the problem, and then submit an agreed and tested patch to those of us upstream; please don't make it for those upstream to pick one of your patches as you are at present.

Quoted 6 lines

Apologies for creating this confusion. I posted a patch and dropped it into the patch system when I found the bug.

However, only when Florian asked about a 'fixes' tag, I went back to the history, and realized that the issue was introduced by Mike during the most recent merge window, and affects xtensa as well.

Quoted 6 lines

So why does Mike's patch have:

in it? It seems you haven't been directly involved in Mike's patch.

Because I cc'ed him on the discussion following the patch that is now in your patch system. So he took that patch and modified it, but retained the original S-o-b and authorship.

Quoted 10 lines

There's something /really/ not right with the process behind this patch.

Quoted 2 lines

I don't have a preference which patch gets applied, though, so please indicate your preference, and we will adapt accordingly.

I asked for you both to come to a concensus about how you want to proceed, and now you're throwing it back on to me to solve your(pl) mis-communication issue. We haven't heard from Mike yet.

I am not throwing it back to you. I merely indicated that I have no preference, and since Mike is the one who introduced this issue in the first place, I am expecting him to drive this. And indeed, we haven't heard from him yet.

Quoted 4 lines

Clearly, I wasn't blunt and stroppy enough to be properly understood. Sort it out between yourselves and tell me which patch you want me to apply.

I would like you to ack this version of the patch, and disregard the one in the patch system, so that Mike can take this one through the memblock tree where the issue originated in the first place.

#5 lore raw
Quoted 61 lines · Ard Biesheuvel
Quoted 59 lines · Russell King - ARM Linux admin
Quoted 48 lines · Ard Biesheuvel
Quoted 38 lines · Russell King - ARM Linux admin
Quoted 22 lines · Mike Rapoport

From: Ard Biesheuvel <[email protected]>

free_highpages() iterates over the free memblock regions in high memory, and marks each page as available for the memory management system.

Until commit cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages") it rounded beginning of each region upwards and end of each region downwards.

However, after that commit free_highmem() rounds the beginning and end of each region downwards, and we may end up freeing a page that is memblock_reserve()d, resulting in memory corruption.

Restore the original rounding of the region boundaries to avoid freeing reserved pages.

Fixes cddb5ddf2b76 ("arm, xtensa: simplify initialization of high memory pages")
Quoted 8 lines

---

Max, Russell,

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

v2: fix words order in the commit message

I really don't understand what is going on here; there seems to be a total disconnect of communication between yourself and Ard. Ard has already submitted a different patch for this to the patch system already, sent yesterday.

Quoted 7 lines

https://www.armlinux.org.uk/developer/patches/viewpatch.php?id=9021/1

Please discuss between yourselves how you want to solve the problem, and then submit an agreed and tested patch to those of us upstream; please don't make it for those upstream to pick one of your patches as you are at present.

Quoted 6 lines

Apologies for creating this confusion. I posted a patch and dropped it into the patch system when I found the bug.

However, only when Florian asked about a 'fixes' tag, I went back to the history, and realized that the issue was introduced by Mike during the most recent merge window, and affects xtensa as well.

So why does Mike's patch have:

Quoted 8 lines
Quoted 4 lines

in it? It seems you haven't been directly involved in Mike's patch.

Because I cc'ed him on the discussion following the patch that is now in your patch system. So he took that patch and modified it, but retained the original S-o-b and authorship.

Right, that's exactly what happened.

Quoted 11 lines
Quoted 10 lines

There's something /really/ not right with the process behind this patch.

Quoted 2 lines

I don't have a preference which patch gets applied, though, so please indicate your preference, and we will adapt accordingly.

I asked for you both to come to a concensus about how you want to proceed, and now you're throwing it back on to me to solve your(pl) mis-communication issue. We haven't heard from Mike yet.

Quoted 4 lines

I am not throwing it back to you. I merely indicated that I have no preference, and since Mike is the one who introduced this issue in the first place, I am expecting him to drive this. And indeed, we haven't heard from him yet.

I didn't know that Ard's patch was already in the patch system. I took the patch from the list, updated it, added a fix for xtensa and resend while retaining Ard's authourship and s-o-b.

Quoted 8 lines
Quoted 4 lines

Clearly, I wasn't blunt and stroppy enough to be properly understood. Sort it out between yourselves and tell me which patch you want me to apply.

I would like you to ack this version of the patch, and disregard the one in the patch system, so that Mike can take this one through the memblock tree where the issue originated in the first place.

Agree.

Signature · 2 lines
Sincerely yours,
Mike.
maintainer
#7 lore raw
In reply to Mike Rapoport
Quoted 2 lines · Mike Rapoport

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

Going through the memblock tree sounds right to me.

Signature · 2 lines
Thanks.
-- Max
#8 lore raw
Quoted 5 lines · Max Filippov
Quoted 2 lines · Mike Rapoport

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

Going through the memblock tree sounds right to me.

Can I treat this as Ack?

Quoted 3 lines
Signature · 2 lines
Thanks.
-- Max
Signature · 2 lines
Sincerely yours,
Mike.
maintainer Acked-by
#9 lore raw
Quoted 9 lines · Mike Rapoport
Quoted 5 lines · Max Filippov
Quoted 2 lines · Mike Rapoport

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

Going through the memblock tree sounds right to me.

Can I treat this as Ack?

Sure, for the xtensa part:

Signature · 2 lines
Thanks.
-- Max
#10 lore raw
Quoted 8 lines · Max Filippov
Quoted 6 lines · Mike Rapoport
Quoted 4 lines · Max Filippov
Quoted 2 lines · Mike Rapoport

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

Quoted 7 lines
Quoted 3 lines
Quoted 1 line

Going through the memblock tree sounds right to me.

Can I treat this as Ack?

Sure, for the xtensa part:

Could we get this queued up please?

#11 lore raw
Quoted 11 lines · Ard Biesheuvel
Quoted 10 lines · Max Filippov
Quoted 8 lines · Mike Rapoport
Quoted 5 lines · Max Filippov
Quoted 2 lines · Mike Rapoport

Please let me know how do you prefer to take it upstream. If needed this can go via memblock tree.

Going through the memblock tree sounds right to me.

Quoted 7 lines
Quoted 5 lines
Quoted 1 line

Can I treat this as Ack?

Sure, for the xtensa part:

Could we get this queued up please?

It's in memblock/fixes now, I'd like to have it in next for day or two.

Signature · 2 lines
Sincerely yours,
Mike.
#6 lore raw
In reply to Mike Rapoport

Hi Russell,

Quoted 4 lines · Mike Rapoport
Quoted 3 lines · Ard Biesheuvel
Quoted 1 line · Russell King - ARM Linux admin
Quoted 8 lines
Quoted 8 lines
Quoted 4 lines

Clearly, I wasn't blunt and stroppy enough to be properly understood. Sort it out between yourselves and tell me which patch you want me to apply.

I would like you to ack this version of the patch, and disregard the one in the patch system, so that Mike can take this one through the memblock tree where the issue originated in the first place.

Can I please have your ack and move forward with pushing this via memblock tree?

Signature · 2 lines
Sincerely yours,
Mike.