madvise.2: Document MADV_POPULATE_READ and MADV_POPULATE_WRITE
Subsystems (3)
PATCH v2 1/1 madvise.2: Document MADV_POPULATE_READ and MADV_POPULATE_WRITE
MADV_POPULATE_READ and MADV_POPULATE_WRITE have been merged into upstream Linux via commit 4ca9b3859dac ("mm/madvise: introduce MADV_POPULATE_(READ|WRITE) to prefault page tables"), part of v5.14-rc1.
Further, commit eb2faa513c24 ("mm/madvise: report SIGBUS as -EFAULT for MADV_POPULATE_(READ|WRITE)"), part of v5.14-rc6, made sure that SIGBUS is converted to -EFAULT instead of -EINVAL.
Let's document the behavior and error conditions of these new madvise() options.
---
v1 -> v2:
- Use semantic newlines in all cases
- Add two missing "
- Document -EFAULT handling
- Rephrase some parts to make it more generic: VM_PFNMAP and VM_IO are only examples for special mappings
--- man2/madvise.2 | 107 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 107 insertions(+)
man2/madvise.2 +107 -0
@@ -469,6 +469,72 @@ If a page is file-backed and dirty, it will be written back to the backing storage. The advice might be ignored for some pages in the range when it is not applicable. +.TP +.BR MADV_POPULATE_READ " (since Linux 5.14)" +Populate (prefault) page tables readable for the whole range without actually +reading memory. +Depending on the underlying mapping, +map the shared zeropage, +preallocate memory or read the underlying file; +files with holes might or might not preallocate blocks. +Do not generate +.B SIGBUS +when populating fails, +return an error instead. +.IP +If +.B MADV_POPULATE_READ +succeeds, +all page tables have been populated (prefaulted) readable once. +If +.B MADV_POPULATE_READ +fails, +some page tables might have been populated. +.IP +.B MADV_POPULATE_READ +cannot be applied to mappings without read permissions +and special mappings, +for example, +marked with the kernel-internal +.B VM_PFNMAP +and +.BR VM_IO . +.IP +Note that with +.BR MADV_POPULATE_READ , +the process can be killed at any moment when the system runs out of memory. +.TP +.BR MADV_POPULATE_WRITE " (since Linux 5.14)" +Populate (prefault) page tables writable for the whole range without actually +writing memory. +Depending on the underlying mapping, +preallocate memory or read the underlying file; +files with holes will preallocate blocks. +Do not generate +.B SIGBUS +when populating fails, +return an error instead. +.IP +If +.B MADV_POPULATE_WRITE +succeeds, +all page tables have been populated (prefaulted) writable once. +If +.B MADV_POPULATE_WRITE +fails, some page tables might have been populated. +.IP +.B MADV_POPULATE_WRITE +cannot be applied to mappings without write permissions +and special mappings, +for example, +marked with the kernel-internal +.B VM_PFNMAP +and +.BR VM_IO . +.IP +Note that with +.BR MADV_POPULATE_WRITE , +the process can be killed at any moment when the system runs out of memory. .SH RETURN VALUE On success, .BR madvise ()
@@ -490,6 +556,17 @@ A kernel resource was temporarily unavailable. .B EBADF The map exists, but the area maps something that isn't a file. .TP +.B EFAULT +.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE , +and populating (prefaulting) page tables failed because a +.B SIGBUS +would have been generated on actual memory access and the reason is not a +HW poisoned page. +.TP .B EINVAL .I addr is not page-aligned or
@@ -533,6 +610,18 @@ or .BR VM_PFNMAP ranges. .TP +.B EINVAL +.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE , +but the specified address range includes ranges with insufficient permissions +or incompatible mappings such as +.B VM_IO +or +.BR VM_PFNMAP. +.TP .B EIO (for .BR MADV_WILLNEED )
@@ -548,6 +637,15 @@ Not enough memory: paging in failed. Addresses in the specified range are not currently mapped, or are outside the address space of the process. .TP +.B ENOMEM +.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE , +and populating (prefaulting) page tables failed because there was not enough +memory. +.TP .B EPERM .I advice is
@@ -555,6 +653,15 @@ is but the caller does not have the .B CAP_SYS_ADMIN capability. +.TP +.B EHWPOISON +.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE , +and populating (prefaulting) page tables failed because a HW poisoned page +was encountered. .SH VERSIONS Since Linux 3.18, .\" commit d3ac21cacc24790eb45d735769f35753f5b56ceb
Signature · 1 line
2.31.1
Hello David,
Thank you for writing this! Could you please take
a look at the comments below and revise?
Quoted 40 lines · David Hildenbrand
MADV_POPULATE_READandMADV_POPULATE_WRITEhave been merged into upstream Linux via commit4ca9b3859dac("mm/madvise: introduceMADV_POPULATE_(READ|WRITE)to prefault page tables"), part of v5.14-rc1.Further, commit
eb2faa513c24("mm/madvise: report SIGBUS as-EFAULTforMADV_POPULATE_(READ|WRITE)"), part of v5.14-rc6, made sure that SIGBUS is converted to-EFAULTinstead of-EINVAL.Let's document the behavior and error conditions of these new
madvise()options.Acked-by Pankaj Gupta <[email protected]>Cc Alejandro Colomar <[email protected]>Cc Michael Kerrisk <[email protected]>Cc Andrew Morton <[email protected]>Cc Michal Hocko <[email protected]>Cc Oscar Salvador <[email protected]>Cc Jann Horn <[email protected]>Cc Mike Rapoport <[email protected]>Cc Linux API <[email protected]>Signed-off-by David Hildenbrand <[email protected]>---
v1 -> v2:
- Use semantic newlines in all cases
- Add two missing "
- Document-EFAULThandling
- Rephrase some parts to make it more generic:VM_PFNMAPandVM_IOare only examples for special mappings--- man2/madvise.2 | 107 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 107 insertions(+)man2/madvise.2 +0
-0
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 474
Quoted 6 lines
man2/madvise.2 +3
-0@@ -469,6 +469,72 @@ If a page is file-backed and dirty, it will be written back to the backing storage. The advice might be ignored for some pages in the range when it is not applicable. +.TP +.BR MADV_POPULATE_READ " (since Linux 5.14)" +Populate (prefault) page tables readable for the whole range without actually
I have trouble to understand "Populate (prefault) page tables readable". Does it mean that it is just the page tables are being populated, and the PTEs are marked to indicate that the pages are readable? If yes, I think some rewording would help.
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 475
Quoted 1 line
man2/madvise.2 +1
-0+reading memory.
I don't understand "without actually reading memory"? Do you mean, "without actually faulting in the pages"; or something else?
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 483
Quoted 8 lines
man2/madvise.2 +8
-0+Depending on the underlying mapping, +map the shared zeropage, +preallocate memory or read the underlying file; +files with holes might or might not preallocate blocks. +Do not generate +.B SIGBUS +when populating fails, +return an error instead.
Better:
[[
If populating fails, a
.B SIGBUS
signal is not generated; instead, an error i returned. ]]
Quoted 5 lines
man2/madvise.2 +5
-0+.IP +If +.B MADV_POPULATE_READ +succeeds, +all page tables have been populated (prefaulted) readable once.
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 498
Quoted 10 lines
man2/madvise.2 +10
-0+If +.B MADV_POPULATE_READ +fails, +some page tables might have been populated. +.IP +.B MADV_POPULATE_READ +cannot be applied to mappings without read permissions +and special mappings, +for example, +marked with the kernel-internal
s/marked/mappings marked/
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 500
Quoted 2 lines
man2/madvise.2 +2
-0+.B VM_PFNMAP +and
Just checking: should it be "and" or "or" here"?
Looking at the EINVAL error below, I guess "or", and a better wording would be:
[[
...for example, mappings marked with kernel-internal flags such as .B VMPPFNMAP
or
.BR BR_V_IO.
]]
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 508
Quoted 8 lines
man2/madvise.2 +8
-0+.BR VM_IO . +.IP +Note that with +.BR MADV_POPULATE_READ , +the process can be killed at any moment when the system runs out of memory. +.TP +.BR MADV_POPULATE_WRITE " (since Linux 5.14)" +Populate (prefault) page tables writable for the whole range without actually
I have trouble to understand "Populate (prefault) page tables writable". Does it mean that it is just the page tables are being populated, and the PTEs are marked to indicate that the pages are writable? If yes, I think some rewording would help.
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 509
Quoted 1 line
man2/madvise.2 +1
-0+writing memory.
I don't understand "without actually writing memory"? Do you mean, "without actually faulting in the pages"; or something else?
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 516
Quoted 7 lines
man2/madvise.2 +7
-0+Depending on the underlying mapping, +preallocate memory or read the underlying file; +files with holes will preallocate blocks. +Do not generate +.B SIGBUS +when populating fails, +return an error instead.
Better:
[[
If populating fails, a
.B SIGBUS
signal is not generated; instead, an error i returned. ]]
+.IP
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 530
Quoted 13 lines
man2/madvise.2 +13
-0+If +.B MADV_POPULATE_WRITE +succeeds, +all page tables have been populated (prefaulted) writable once. +If +.B MADV_POPULATE_WRITE +fails, some page tables might have been populated. +.IP +.B MADV_POPULATE_WRITE +cannot be applied to mappings without write permissions +and special mappings, +for example, +marked with the kernel-internal
s/marked/mappings marked/
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 531
Quoted 2 lines
man2/madvise.2 +2
-0+.B VM_PFNMAP +and
Just checking: should it be "and" or "or" here"?
Looking at the EINVAL error below, I guess "or", and a better wording would be:
[[
...for example, mappings marked with kernel-internal flags such as .B VMPPFNMAP
or
.BR BR_V_IO.
]]
Quoted 12 lines
man2/madvise.2 +5
-0+.BR VM_IO . +.IP +Note that with +.BR MADV_POPULATE_WRITE , +the process can be killed at any moment when the system runs out of memory. .SH RETURN VALUE On success, .BR madvise ()@@ -490,6 +556,17 @@ A kernel resource was temporarily unavailable. .B EBADF The map exists, but the area maps something that isn't a file. .TP
man2/madvise.2 A kernel resource was temporarily unavailable. · line 568
Quoted 10 lines
man2/madvise.2 +10
-0+.B EFAULT +.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE , +and populating (prefaulting) page tables failed because a +.B SIGBUS +would have been generated on actual memory access and the reason is not a +HW poisoned page.
Maybe:
s/.$/(see the description of MADV_HWPOISON in this page)./ ?
Quoted 9 lines
man2/madvise.2 +2
-0+.TP .B EINVAL .I addr is not page-aligned or@@ -533,6 +610,18 @@ or .BR VM_PFNMAP ranges. .TP +.B EINVAL
man2/madvise.2 or · line 623
Quoted 10 lines
man2/madvise.2 +10
-0+.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE , +but the specified address range includes ranges with insufficient permissions +or incompatible mappings such as +.B VM_IO +or +.BR VM_PFNMAP.
s/.BR VM_PFNMAP./.BR VM_PFNMAP ./
Quoted 26 lines
man2/madvise.2 +12
-0+.TP .B EIO (for .BR MADV_WILLNEED )@@ -548,6 +637,15 @@ Not enough memory: paging in failed. Addresses in the specified range are not currently mapped, or are outside the address space of the process. .TP +.B ENOMEM +.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE , +and populating (prefaulting) page tables failed because there was not enough +memory. +.TP .B EPERM .I advice is@@ -555,6 +653,15 @@ is but the caller does not have the .B CAP_SYS_ADMIN capability. +.TP +.B EHWPOISON
man2/madvise.2 is · line 667
Quoted 10 lines
man2/madvise.2 +7
-0+.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE , +and populating (prefaulting) page tables failed because a HW poisoned page +was encountered. .SH VERSIONS Since Linux 3.18, .\" commit d3ac21cacc24790eb45d735769f35753f5b56ceb
Thanks,
Michael
Signature · 3 lines
Michael Kerrisk Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/ Linux/UNIX System Programming Training: http://man7.org/training/
Quoted 4 lines · Michael Kerrisk (man-pages)
Hello David,
Thank you for writing this! Could you please take
a look at the comments below and revise?
Hi Michael,
thanks for your valuable input. Your feedback will certainly make this easier to understand for people that are not heavily involved in MM work :)
[...]
Quoted 19 lines
Quoted 8 lines
man2/madvise.2 | 107 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 107 insertions(+)man2/madvise.2 +0
-0man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 474
Quoted 6 lines
man2/madvise.2 +3
-0@@ -469,6 +469,72 @@ If a page is file-backed and dirty, it will be written back to the backing storage. The advice might be ignored for some pages in the range when it is not applicable. +.TP +.BR MADV_POPULATE_READ " (since Linux 5.14)" +Populate (prefault) page tables readable for the whole range without actuallyI have trouble to understand "Populate (prefault) page tables readable". Does it mean that it is just the page tables are being populated, and the PTEs are marked to indicate that the pages are readable? If yes, I think some rewording would help.
I actually tried phrasing it similar to our MAP_POPULATE documentation:
("Populate (prefault) page tables for a mapping.")
We will prefault all pages, faulting them in.
Quoted 5 lines
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 475
Quoted 2 lines
man2/madvise.2 +1
-0+reading memory.I don't understand "without actually reading memory"? Do you mean, "without actually faulting in the pages"; or something else?
"Populate (prefault) page tables readable, faulting in all pages in the range just as if manually reading one byte of each page; however, avoid the actual memory access that would have been performed after handling the fault."
Does that make it clearer? (avoiding eventually touching the page at all can be beneficial, especially when dealing with DAX memory where memory access might be expensive)
Quoted 12 lines
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 483
Quoted 9 lines
man2/madvise.2 +8
-0+Depending on the underlying mapping, +map the shared zeropage, +preallocate memory or read the underlying file; +files with holes might or might not preallocate blocks. +Do not generate +.B SIGBUS +when populating fails, +return an error instead.Better:
Quoted 6 lines
[[
If populating fails, a
.B SIGBUS
signal is not generated; instead, an error i returned. ]]
Sure, thanks.
Quoted 26 lines
Quoted 5 lines
man2/madvise.2 +5
-0+.IP +If +.B MADV_POPULATE_READ +succeeds, +all page tables have been populated (prefaulted) readable once.man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 498
Quoted 10 lines
man2/madvise.2 +10
-0+If +.B MADV_POPULATE_READ +fails, +some page tables might have been populated. +.IP +.B MADV_POPULATE_READ +cannot be applied to mappings without read permissions +and special mappings, +for example, +marked with the kernel-internals/marked/mappings marked/
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 500
Quoted 2 lines
man2/madvise.2 +2
-0+.B VM_PFNMAP +andJust checking: should it be "and" or "or" here"?
Looking at the EINVAL error below, I guess "or", and a better wording would be:
Quoted 6 lines
[[
...for example, mappings marked with kernel-internal flags such as .B VMPPFNMAP
or
.BRBR_V_IO.
]]
Much better. Note that there might be more types of mappings that won't work (e.g., initially also secretmem IIRC).
Quoted 15 lines
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 508
Quoted 9 lines
man2/madvise.2 +8
-0+.BR VM_IO . +.IP +Note that with +.BR MADV_POPULATE_READ , +the process can be killed at any moment when the system runs out of memory. +.TP +.BR MADV_POPULATE_WRITE " (since Linux 5.14)" +Populate (prefault) page tables writable for the whole range without actuallyI have trouble to understand "Populate (prefault) page tables writable". Does it mean that it is just the page tables are being populated, and the PTEs are marked to indicate that the pages are writable? If yes, I think some rewording would help.
Quoted 5 lines
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 509
Quoted 1 line
man2/madvise.2 +1
-0+writing memory.I don't understand "without actually writing memory"? Do you mean, "without actually faulting in the pages"; or something else?
Similar to the other wording:
"Populate (prefault) page tables writable, faulting in all pages in the range just as if manually writing one byte of each page; however, avoid the actual memory access that would have been performed after handling the fault."
Quoted 10 lines
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 516
Quoted 7 lines
man2/madvise.2 +7
-0+Depending on the underlying mapping, +preallocate memory or read the underlying file; +files with holes will preallocate blocks. +Do not generate +.B SIGBUS +when populating fails, +return an error instead.Better:
Quoted 6 lines
[[
If populating fails, a
.B SIGBUS
signal is not generated; instead, an error i returned. ]]
Ack.
Quoted 25 lines
+.IP
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 530
Quoted 13 lines
man2/madvise.2 +13
-0+If +.B MADV_POPULATE_WRITE +succeeds, +all page tables have been populated (prefaulted) writable once. +If +.B MADV_POPULATE_WRITE +fails, some page tables might have been populated. +.IP +.B MADV_POPULATE_WRITE +cannot be applied to mappings without write permissions +and special mappings, +for example, +marked with the kernel-internals/marked/mappings marked/
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 531
Quoted 2 lines
man2/madvise.2 +2
-0+.B VM_PFNMAP +andJust checking: should it be "and" or "or" here"?
Looking at the EINVAL error below, I guess "or", and a better wording would be:
Quoted 7 lines
[[
...for example, mappings marked with kernel-internal flags such as .B VMPPFNMAP
or
.BRBR_V_IO.
]]
Ack.
Quoted 18 lines
Quoted 18 lines
man2/madvise.2 +11
-0+.BR VM_IO . +.IP +Note that with +.BR MADV_POPULATE_WRITE , +the process can be killed at any moment when the system runs out of memory. .SH RETURN VALUE On success, .BR madvise ()@@ -490,6 +556,17 @@ A kernel resource was temporarily unavailable. .B EBADF The map exists, but the area maps something that isn't a file. .TP +.B EFAULT +.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE ,
Quoted 9 lines
man2/madvise.2 A kernel resource was temporarily unavailable. · line 568
Quoted 4 lines
man2/madvise.2 +4
-0+and populating (prefaulting) page tables failed because a +.B SIGBUS +would have been generated on actual memory access and the reason is not a +HW poisoned page.Maybe:
s/.$/(see the description ofMADV_HWPOISONin this page)./ ?
Sure, we can add that. But note that MADV_HWPOISON is just one of many ways to HWpoison a page.
Quoted 13 lines
Quoted 13 lines
man2/madvise.2 +6
-0+.TP .B EINVAL .I addr is not page-aligned or@@ -533,6 +610,18 @@ or .BR VM_PFNMAP ranges. .TP +.B EINVAL +.I advice +is +.B MADV_POPULATE_READ +or
Quoted 9 lines
man2/madvise.2 or · line 623
Quoted 6 lines
man2/madvise.2 +6
-0+.BR MADV_POPULATE_WRITE , +but the specified address range includes ranges with insufficient permissions +or incompatible mappings such as +.B VM_IO +or +.BR VM_PFNMAP.s/.BR VM_PFNMAP./.BR
VM_PFNMAP./
Agreed.
Quoted 30 lines
Quoted 30 lines
man2/madvise.2 +16
-0+.TP .B EIO (for .BR MADV_WILLNEED )@@ -548,6 +637,15 @@ Not enough memory: paging in failed. Addresses in the specified range are not currently mapped, or are outside the address space of the process. .TP +.B ENOMEM +.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE , +and populating (prefaulting) page tables failed because there was not enough +memory. +.TP .B EPERM .I advice is@@ -555,6 +653,15 @@ is but the caller does not have the .B CAP_SYS_ADMIN capability. +.TP +.B EHWPOISON +.I advice +is +.B MADV_POPULATE_READ +or
Quoted 9 lines
man2/madvise.2 is · line 667
Quoted 6 lines
man2/madvise.2 +3
-0+.BR MADV_POPULATE_WRITE , +and populating (prefaulting) page tables failed because a HW poisoned page +was encountered. .SH VERSIONS Since Linux 3.18, .\" commit d3ac21cacc24790eb45d735769f35753f5b56cebThanks,
Thanks a lot Michael!
Signature · 3 lines
Thanks, David / dhildenb
Hello David,
Quoted 33 lines · David Hildenbrand
Quoted 4 lines · Michael Kerrisk (man-pages)
Hello David,
Thank you for writing this! Could you please take
a look at the comments below and revise?Hi Michael,
thanks for your valuable input. Your feedback will certainly make this easier to understand for people that are not heavily involved in MM work :)
[...]
Quoted 19 lines
Quoted 8 lines
man2/madvise.2 | 107 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 107 insertions(+)man2/madvise.2 +0
-0man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 474
Quoted 6 lines
man2/madvise.2 +3
-0@@ -469,6 +469,72 @@ If a page is file-backed and dirty, it will be written back to the backing storage. The advice might be ignored for some pages in the range when it is not applicable. +.TP +.BR MADV_POPULATE_READ " (since Linux 5.14)" +Populate (prefault) page tables readable for the whole range without actuallyI have trouble to understand "Populate (prefault) page tables readable". Does it mean that it is just the page tables are being populated, and the PTEs are marked to indicate that the pages are readable? If yes, I think some rewording would help.
Quoted 2 lines
I actually tried phrasing it similar to our
MAP_POPULATEdocumentation:
("Populate (prefault) page tables for a mapping.")
Yeah, well that description is a bit thin too :-}.
Quoted 7 lines
We will prefault all pages, faulting them in.
Quoted 5 lines
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 475
Quoted 2 lines
man2/madvise.2 +1
-0+reading memory.I don't understand "without actually reading memory"? Do you mean, "without actually faulting in the pages"; or something else?
Quoted 8 lines
"Populate (prefault) page tables readable, faulting in all pages in the range just as if manually reading one byte of each page; however, avoid the actual memory access that would have been performed after handling the fault."
Does that make it clearer? (avoiding eventually touching the page at all can be beneficial, especially when dealing with DAX memory where memory access might be expensive)
That text is much better. But, what's still not clear to me then is the dfference between mmap(2) MAP_POPULATE, and MADV_POPULATE_READ and MADV_POPULATE_WRITE. What is the differnece, and in what situations would one prefer one or the other approach? I think it would be helpful if the manual page said something about these details.
Quoted 53 lines
Quoted 11 lines
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 483
Quoted 8 lines
man2/madvise.2 +8
-0+Depending on the underlying mapping, +map the shared zeropage, +preallocate memory or read the underlying file; +files with holes might or might not preallocate blocks. +Do not generate +.B SIGBUS +when populating fails, +return an error instead.Better:
Quoted 6 lines
[[
If populating fails, a
.B SIGBUS
signal is not generated; instead, an error i returned. ]]Sure, thanks.
Quoted 26 lines
Quoted 5 lines
man2/madvise.2 +5
-0+.IP +If +.B MADV_POPULATE_READ +succeeds, +all page tables have been populated (prefaulted) readable once.man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 498
Quoted 10 lines
man2/madvise.2 +10
-0+If +.B MADV_POPULATE_READ +fails, +some page tables might have been populated. +.IP +.B MADV_POPULATE_READ +cannot be applied to mappings without read permissions +and special mappings, +for example, +marked with the kernel-internals/marked/mappings marked/
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 500
Quoted 2 lines
man2/madvise.2 +2
-0+.B VM_PFNMAP +andJust checking: should it be "and" or "or" here"?
Looking at the EINVAL error below, I guess "or", and a better wording would be:
Quoted 6 lines
[[
...for example, mappings marked with kernel-internal flags such as .B VMPPFNMAP
or
.BRBR_V_IO.
]]
Quoted 2 lines
Much better. Note that there might be more types of mappings that won't work (e.g., initially also secretmem IIRC).
Ahh nice. Since there's about to be a memfd_secret() manual page, I suggest adding also "or secret memory regions created using memfd_secret(2)".
Quoted 20 lines
Quoted 14 lines
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 508
Quoted 8 lines
man2/madvise.2 +8
-0+.BR VM_IO . +.IP +Note that with +.BR MADV_POPULATE_READ , +the process can be killed at any moment when the system runs out of memory. +.TP +.BR MADV_POPULATE_WRITE " (since Linux 5.14)" +Populate (prefault) page tables writable for the whole range without actuallyI have trouble to understand "Populate (prefault) page tables writable". Does it mean that it is just the page tables are being populated, and the PTEs are marked to indicate that the pages are writable? If yes, I think some rewording would help.
Quoted 5 lines
man2/madvise.2 If a page is file-backed and dirty, it will be written back to the backing · line 509
Quoted 1 line
man2/madvise.2 +1
-0+writing memory.I don't understand "without actually writing memory"? Do you mean, "without actually faulting in the pages"; or something else?
Quoted 6 lines
Similar to the other wording:
"Populate (prefault) page tables writable, faulting in all pages in the range just as if manually writing one byte of each page; however, avoid the actual memory access that would have been performed after handling the fault."
Much better, but see also my comments above re MADV_POPULATE_READ.
[...]
Quoted 16 lines
Quoted 6 lines
Quoted 6 lines
man2/madvise.2 +6
-0+.B EFAULT +.I advice +is +.B MADV_POPULATE_READ +or +.BR MADV_POPULATE_WRITE ,Quoted 9 lines
man2/madvise.2 A kernel resource was temporarily unavailable. · line 568
Quoted 4 lines
man2/madvise.2 +4
-0+and populating (prefaulting) page tables failed because a +.B SIGBUS +would have been generated on actual memory access and the reason is not a +HW poisoned page.Maybe:
s/.$/(see the description ofMADV_HWPOISONin this page)./ ?
Quoted 2 lines
Sure, we can add that. But note that
MADV_HWPOISONis just one of many ways to HWpoison a page.
Then maybe something like:
"(HW poisoned pages can, for example, be created using the MADV_HWPOISON flag described elsewhere in this page.)"
[...]
Thanks,
Michael
Signature · 3 lines
Michael Kerrisk Linux man-pages maintainer; http://www.kernel.org/doc/man-pages/ Linux/UNIX System Programming Training: http://man7.org/training/
Quoted 12 lines
Quoted 11 lines
Quoted 2 lines
I don't understand "without actually reading memory"? Do you mean, "without actually faulting in the pages"; or something else?
"Populate (prefault) page tables readable, faulting in all pages in the range just as if manually reading one byte of each page; however, avoid the actual memory access that would have been performed after handling the fault."
Does that make it clearer? (avoiding eventually touching the page at all can be beneficial, especially when dealing with DAX memory where memory access might be expensive)
Quoted 5 lines
That text is much better. But, what's still not clear to me then is the dfference between mmap(2)
MAP_POPULATE, andMADV_POPULATE_READandMADV_POPULATE_WRITE. What is the differnece, and in what situations would one prefer one or the other approach? I think it would be helpful if the manual page said something about these details.
Well, MAP_POPULATE
1. Can only be used on new mappings, not on existing mappings and especially not on parts of (sparse) memory mappings.
2. Hides actual population errors, simply returning "success"
3. Cannot specify the actual faultin behavior (readable vs. writable). Private mappings are always faulted in writable, shared mappings writable.
MADV_POPULATE_WRITE is the way to go to preallocate memory, especially hugetlbfs. MADV_POPULATE_READ can be used to just populate the shared zero page, or to fault-in file-backed memory without marking everything dirty such that it won't have to be written back to disk.
for MADV_POPULATE_READ
"In contrast to MAP_POPULATE, MADV_POPULATE_READ does not hide errors, can be applied to (parts of) existing mappings and will always populate (prefault) page tables readable. One example use case is prefaulting a file mapping, reading all file content from disk; however, pages won't be dirtied and consequently won't have to be written back to disk when evicting the pages from memory."
Suggestions welcome :)
for MADV_POPULATE_WRITE
"In contrast to MAP_POPULATE, MADV_POPULATE_WRITE does not hide errors, can be applied to (parts of) existing mappings and will always populate (prefault) page tables writable. One example use case is preallocating memory, breaking any CoW (Copy on Write)."
Again, suggestions welcome :)
[...]
Quoted 6 lines
Quoted 2 lines
Much better. Note that there might be more types of mappings that won't work (e.g., initially also secretmem IIRC).
Ahh nice. Since there's about to be a
memfd_secret()manual page, I suggest adding also "or secret memory regions created usingmemfd_secret(2)".
Can do, thanks!
[...]
Quoted 8 lines
Quoted 3 lines
Sure, we can add that. But note that
MADV_HWPOISONis just one of many ways to HWpoison a page.Then maybe something like:
"(HW poisoned pages can, for example, be created using the
MADV_HWPOISONflag described elsewhere in this page.)"
Makes sense, I'll reuse the same description at the other place in this file where I mention HW poisoned pages.
Signature · 3 lines
Thanks, David / dhildenb