Skip to contentSkip to conversation

Documentation/vm: Rework and extend highmem.rst

PATCH In review Fabio M. De Francesco 2 participants 7 messages Apr 15, 2022 – Apr 21, 2022 Documentation

lore
Share & tools
lore
#0 lore raw

v3 cover Documentation/vm: Rework and extend highmem.rst

Extend and rework the "Temporary Virtual Mappings" section of the highmem.rst documentation.

Do a partial rework of the paragraph related to kmap() and add a new paragraph in order to document the set of kmap_local_*() functions. Re-order paragraphs in decreasing order of preference of usage.

Include kerneldoc comments from include/linux/highmem.h and remove redundant and obsolete section about kmap_atomic().

v1-v2:
1/2 - According to comments on v1 by Matthew Wilcox and Ira Weiny,
correct a mistake in text, format paragraphs to stay within the 75 characters limit, re-order the flow of the same paragraphs in decreasing order of preference of usage.
2/2 - No patch.

v2->v3:
1/2 - No changes.
2/2 - Added patch to include kernel-doc to highmem.rst and
remove the now redundant section about kmap_atomic().

Fabio M. De Francesco (2):
Documentation/vm: Extend "Temporary Virtual Mappings" section in
highmem.rst
Documentation/vm: Include kernel-doc to highmem.rst

 Documentation/vm/highmem.rst | 96 +++++++++++++++++++++---------------
 1 file changed, 56 insertions(+), 40 deletions(-)
Signature · 1 line
2.34.1
#1 lore raw

PATCH v3 1/2 Documentation/vm: Extend "Temporary Virtual Mappings" section

Extend and rework the "Temporary Virtual Mappings" section of the highmem.rst documentation.

Do a partial rework of the paragraph related to kmap() and add a new paragraph in order to document the set of kmap_local_*() functions. Re-order paragraphs in decreasing order of preference of usage.

Despite the local kmaps were introduced by Thomas Gleixner in October 2020, documentation was still missing information about them. These additions rely largely on Gleixner's patches, Jonathan Corbet's LWN articles, comments by Ira Weiny and Matthew Wilcox, and in-code comments from ./include/linux/highmem.h.

Cc Jonathan Corbet <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Ira Weiny <[email protected]>
Cc Matthew Wilcox <[email protected]>
Cc Peter Zijlstra <[email protected]>
---
 Documentation/vm/highmem.rst | 67 ++++++++++++++++++++++++++++++------
 1 file changed, 56 insertions(+), 11 deletions(-)
Documentation/vm/highmem.rst +56 -11
index 0f69a9fec34d..12dcfbee094d 100644
@@ -52,24 +52,69 @@ Temporary Virtual Mappings
 
 The kernel contains several ways of creating temporary mappings:
 
-* vmap().  This can be used to make a long duration mapping of multiple
-  physical pages into a contiguous virtual space.  It needs global
-  synchronization to unmap.
-
-* kmap().  This permits a short duration mapping of a single page.  It needs
-  global synchronization, but is amortized somewhat.  It is also prone to
-  deadlocks when using in a nested fashion, and so it is not recommended for
-  new code.
+* kmap_local_*().  These provide a set of functions that are used to require
+  short term mappings. They can be invoked from any context (including
+  interrupts) but the mappings can only be used in the context which acquired
+  it.
+
+  These mappings are per thread, CPU local (i.e., migration from one CPU to
+  another is disabled - this is why they are called "local"), but they don't
+  disable preemption. It's valid to take pagefaults in a local kmap region,
+  unless the context in which the local mapping is acquired does not allow
+  it for other reasons.
+
+  If a task holding local kmaps is preempted, the maps are removed on context
+  switch and restored when the task comes back on the CPU. As the maps are
+  strictly CPU local, it is guaranteed that the task stays on the CPU and
+  that the CPU cannot be unplugged until the local kmaps are released.
+
+  Nesting kmap_local.*() and kmap_atomic.*() mappings is allowed to a certain
+  extent (up to KMAP_TYPE_NR). Nested kmap_local.*() and kunmap_local.*()
+  invocations have to be strictly ordered because the map implementation is
+  stack based.
 
 * kmap_atomic().  This permits a very short duration mapping of a single
   page.  Since the mapping is restricted to the CPU that issued it, it
   performs well, but the issuing task is therefore required to stay on that
   CPU until it has finished, lest some other task displace its mappings.
 
-  kmap_atomic() may also be used by interrupt contexts, since it is does not
-  sleep and the caller may not sleep until after kunmap_atomic() is called.
+  kmap_atomic() may also be used by interrupt contexts, since it does not
+  sleep and the caller too can not sleep until after kunmap_atomic() is called.
+  Each call of kmap_atomic() in the kernel creates a non-preemptible section and
+  disable pagefaults.
+
+  This could be a source of unwanted latency, so it should be only used if it is
+  absolutely required, otherwise the corresponding kmap_local_*() variant should
+  be used if it is feasible.
+
+  On 64-bit systems, calls to kmap() and kmap_atomic() have no real work to do
+  because a 64-bit address space is more than sufficient to address all the
+  physical memory, so all of physical memory appears in the direct mapping.
+
+  It is assumed that k[un]map_atomic() won't fail.
 
-  It may be assumed that k[un]map_atomic() won't fail.
+* kmap().  This should be used to make short duration mapping of a single
+  page with no restrictions on preemption or migration. It comes with an
+  overhead as mapping space is restricted and protected by a global lock
+  for synchronization. When mapping is no more needed, the address that
+  the page was mapped to must be released with kunmap().
+
+  Mapping changes must be propagated across all the CPUs. kmap() also
+  requires global TLB invalidation when the kmap's pool wraps and it might
+  block when the mapping space is fully utilized until a slot becomes
+  available. Therefore, kmap() is only callable from preemptible context.
+
+  All the above work is necessary if a mapping must last for a relatively
+  long time but the bulk of high-memory mappings in the kernel are
+  short-lived and only used in one place. This means that the cost of
+  kmap() is mostly wasted in such cases. kmap() was not intended for long
+  term mappings but it has morphed in that direction and its use is
+  strongly discouraged in newer code and the set of the preceding functions
+  should be preferred.
+
+* vmap().  This can be used to make a long duration mapping of multiple
+  physical pages into a contiguous virtual space.  It needs global
+  synchronization to unmap.
 
 
 Using kmap_atomic
Signature · 1 line
2.34.1
#2 lore raw

Re: [PATCH v3 1/2] Documentation/vm: Extend "Temporary Virtual Mappings" section

Quoted 46 lines · Fabio M. De Francesco

Extend and rework the "Temporary Virtual Mappings" section of the highmem.rst documentation.

Do a partial rework of the paragraph related to kmap() and add a new paragraph in order to document the set of kmap_local_*() functions. Re-order paragraphs in decreasing order of preference of usage.

Despite the local kmaps were introduced by Thomas Gleixner in October 2020, documentation was still missing information about them. These additions rely largely on Gleixner's patches, Jonathan Corbet's LWN articles, comments by Ira Weiny and Matthew Wilcox, and in-code comments from ./include/linux/highmem.h.

Cc Jonathan Corbet <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Ira Weiny <[email protected]>
Cc Matthew Wilcox <[email protected]>
Cc Peter Zijlstra <[email protected]>
---
 Documentation/vm/highmem.rst | 67 ++++++++++++++++++++++++++++++------
 1 file changed, 56 insertions(+), 11 deletions(-)
Documentation/vm/highmem.rst +7 -8
index 0f69a9fec34d..12dcfbee094d 100644
@@ -52,24 +52,69 @@ Temporary Virtual Mappings
 
 The kernel contains several ways of creating temporary mappings:
 
-* vmap().  This can be used to make a long duration mapping of multiple
-  physical pages into a contiguous virtual space.  It needs global
-  synchronization to unmap.
-
-* kmap().  This permits a short duration mapping of a single page.  It needs
-  global synchronization, but is amortized somewhat.  It is also prone to
-  deadlocks when using in a nested fashion, and so it is not recommended for
-  new code.
+* kmap_local_*().  These provide a set of functions that are used to require
+  short term mappings. They can be invoked from any context (including
+  interrupts) but the mappings can only be used in the context which acquired
+  it.
+
+  These mappings are per thread, CPU local (i.e., migration from one CPU to
+  another is disabled - this is why they are called "local"), but they don't

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 71

Quoted 10 lines
Documentation/vm/highmem.rst +10 -0
+  disable preemption. It's valid to take pagefaults in a local kmap region,
+  unless the context in which the local mapping is acquired does not allow
+  it for other reasons.
+
+  If a task holding local kmaps is preempted, the maps are removed on context
+  switch and restored when the task comes back on the CPU. As the maps are
+  strictly CPU local, it is guaranteed that the task stays on the CPU and
+  that the CPU cannot be unplugged until the local kmaps are released.
+
+  Nesting kmap_local.*() and kmap_atomic.*() mappings is allowed to a certain

NIT: Throughout this doc why say 'kmap_local.*'? There is only one call named kmap_local_page().

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 74

Quoted 3 lines
Documentation/vm/highmem.rst +3 -0
+  extent (up to KMAP_TYPE_NR). Nested kmap_local.*() and kunmap_local.*()
+  invocations have to be strictly ordered because the map implementation is
+  stack based.

This type of documentation should (and I believe is, in the kdoc of kmap_local_page(). Why not just add this text (or enhance what is there) and include that here?

Ah I see that patch 2/2 does add the kdocs for the functions... ah ok :-/

Perhaps this section should focus on why to use each of the kmap calls and not how? Leave the how to the kdoc? Although all this information would be nice inside the header for programmers who are looking at using these functions.

Here is an example of how I dealt with this on a recent auxiliary bus documentation update:

https://lore.kernel.org/lkml/[email protected]/

Quoted 8 lines
Documentation/vm/highmem.rst +0 -2
 * kmap_atomic().  This permits a very short duration mapping of a single
   page.  Since the mapping is restricted to the CPU that issued it, it
   performs well, but the issuing task is therefore required to stay on that
   CPU until it has finished, lest some other task displace its mappings.
 
-  kmap_atomic() may also be used by interrupt contexts, since it is does not
-  sleep and the caller may not sleep until after kunmap_atomic() is called.

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 90

Quoted 10 lines
Documentation/vm/highmem.rst +10 -0
+  kmap_atomic() may also be used by interrupt contexts, since it does not
+  sleep and the caller too can not sleep until after kunmap_atomic() is called.
+  Each call of kmap_atomic() in the kernel creates a non-preemptible section and
+  disable pagefaults.
+
+  This could be a source of unwanted latency, so it should be only used if it is
+  absolutely required, otherwise the corresponding kmap_local_*() variant should
+  be used if it is feasible.
+
+  On 64-bit systems, calls to kmap() and kmap_atomic() have no real work to do
                                 ^^^^^^^^^^^^^^^^^^^^^^^^

What about kmap_local_page()?

Quoted 14 lines
Documentation/vm/highmem.rst +12 -1
+  because a 64-bit address space is more than sufficient to address all the
+  physical memory, so all of physical memory appears in the direct mapping.
+
+  It is assumed that k[un]map_atomic() won't fail.
 
-  It may be assumed that k[un]map_atomic() won't fail.
+* kmap().  This should be used to make short duration mapping of a single
+  page with no restrictions on preemption or migration. It comes with an
+  overhead as mapping space is restricted and protected by a global lock
+  for synchronization. When mapping is no more needed, the address that
+  the page was mapped to must be released with kunmap().
+
+  Mapping changes must be propagated across all the CPUs. kmap() also
+  requires global TLB invalidation when the kmap's pool wraps and it might

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 113

Quoted 10 lines
Documentation/vm/highmem.rst +10 -0
+  block when the mapping space is fully utilized until a slot becomes
+  available. Therefore, kmap() is only callable from preemptible context.
+
+  All the above work is necessary if a mapping must last for a relatively
+  long time but the bulk of high-memory mappings in the kernel are
+  short-lived and only used in one place. This means that the cost of
+  kmap() is mostly wasted in such cases. kmap() was not intended for long
+  term mappings but it has morphed in that direction and its use is
+  strongly discouraged in newer code and the set of the preceding functions
+  should be preferred.

I like this paragraph especially the last sentence! Nicely put!

But this is another reason I think this text in the code next to the kmap calls. Then it is more readily available to programmers who are looking at the functions.

Thanks for the great work so far!
Ira

Quoted 10 lines
Documentation/vm/highmem.rst +4 -0
+
+* vmap().  This can be used to make a long duration mapping of multiple
+  physical pages into a contiguous virtual space.  It needs global
+  synchronization to unmap.
 
 
 Using kmap_atomic
Signature · 1 line
2.34.1
#3 lore raw

Re: [PATCH v3 1/2] Documentation/vm: Extend "Temporary Virtual Mappings" section

Quoted 2 lines · On lunedì 18 aprile 2022 23:30:12 CEST Ira Weiny wrote:
Quoted 1 line · Fabio M. De Francesco

Extend and rework the "Temporary Virtual Mappings" section of the

highmem.rst

Quoted 3 lines
Quoted 3 lines

documentation.

Do a partial rework of the paragraph related to kmap() and add a new

paragraph

Quoted 1 line
Quoted 1 line

in order to document the set of kmap_local_*() functions. Re-order

paragraphs

Quoted 3 lines
Quoted 3 lines

in decreasing order of preference of usage.

Despite the local kmaps were introduced by Thomas Gleixner in October

2020,

Quoted 1 line
Quoted 1 line

documentation was still missing information about them. These additions

rely

Quoted 1 line
Quoted 1 line

largely on Gleixner's patches, Jonathan Corbet's LWN articles, comments

by

Quoted 13 lines
Quoted 13 lines

Ira Weiny and Matthew Wilcox, and in-code comments from ./include/linux/highmem.h.

Cc Jonathan Corbet <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Ira Weiny <[email protected]>
Cc Matthew Wilcox <[email protected]>
Cc Peter Zijlstra <[email protected]>
---
 Documentation/vm/highmem.rst | 67 ++++++++++++++++++++++++++++++------
 1 file changed, 56 insertions(+), 11 deletions(-)
Quoted 1 line

Documentation/vm/

Quoted 1 line
Documentation/vm/ renamed from Documentation/vm/highmem.rst +0 -0

highmem.rst

Quoted 8 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 55

Quoted 8 lines

index 0f69a9fec34d..12dcfbee094d 100644

Documentation/vm/highmem.rst +0 -1
@@ -52,24 +52,69 @@ Temporary Virtual Mappings
 
 The kernel contains several ways of creating temporary mappings:
 
-* vmap().  This can be used to make a long duration mapping of 

multiple

Quoted 4 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings

Quoted 4 lines
Documentation/vm/highmem.rst +0 -4
-  physical pages into a contiguous virtual space.  It needs global
-  synchronization to unmap.
-
-* kmap().  This permits a short duration mapping of a single page.  It 

needs

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 60

Quoted 1 line
Documentation/vm/highmem.rst +0 -1
-  global synchronization, but is amortized somewhat.  It is also prone 

to

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 61

Quoted 1 line
Documentation/vm/highmem.rst +0 -1
-  deadlocks when using in a nested fashion, and so it is not 

recommended for

Quoted 2 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 55

Quoted 2 lines
Documentation/vm/highmem.rst +1 -1
-  new code.
+* kmap_local_*().  These provide a set of functions that are used to 

require

Quoted 2 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 57

Quoted 2 lines
Documentation/vm/highmem.rst +2 -0
+  short term mappings. They can be invoked from any context (including
+  interrupts) but the mappings can only be used in the context which 

acquired

Quoted 3 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings

Quoted 3 lines
Documentation/vm/highmem.rst +3 -0
+  it.
+
+  These mappings are per thread, CPU local (i.e., migration from one 

CPU to

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 61

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  another is disabled - this is why they are called "local"), but they 

don't

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 62

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  disable preemption. It's valid to take pagefaults in a local kmap 

region,

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 63

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  unless the context in which the local mapping is acquired does not 

allow

Quoted 3 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings

Quoted 3 lines
Documentation/vm/highmem.rst +3 -0
+  it for other reasons.
+
+  If a task holding local kmaps is preempted, the maps are removed on 

context

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 67

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  switch and restored when the task comes back on the CPU. As the maps 

are

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 68

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  strictly CPU local, it is guaranteed that the task stays on the CPU 

and

Quoted 3 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings

Quoted 3 lines
Documentation/vm/highmem.rst +3 -0
+  that the CPU cannot be unplugged until the local kmaps are released.
+
+  Nesting kmap_local.*() and kmap_atomic.*() mappings is allowed to a 

certain

Quoted 2 lines

NIT: Throughout this doc why say 'kmap_local.*'? There is only one call

named

Quoted 1 line

kmap_local_page().

Let me explain why throughout this text I used "kmap_local.*()"... As I wrote, I relied largely also on Thomas Gleixner's series. He used that way of referring to local kmaps.

I think that this was probably due to his introduction of several local kmaps related functions: kmap_local_page(), kmap_local_page_prot(), kmap_local_pfn().

However, I understand your "NIT" and I'll use a plain kmap_local_page() in the next version of this patch /series. After all, this paragraph is only talking about kmap_local_page().

Quoted 2 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 72

Quoted 2 lines
Documentation/vm/highmem.rst +1 -0
+  extent (up to KMAP_TYPE_NR). Nested kmap_local.*() and 

kunmap_local.*()

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 73

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  invocations have to be strictly ordered because the map 

implementation is

Quoted 4 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 74

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  stack based.

This type of documentation should (and I believe is, in the kdoc of kmap_local_page(). Why not just add this text (or enhance what is there)

and

Quoted 3 lines

include that here?

Ah I see that patch 2/2 does add the kdocs for the functions... ah ok

:-/

Quoted 2 lines

Perhaps this section should focus on why to use each of the kmap calls

and not

Quoted 1 line

how? Leave the how to the kdoc? Although all this information would be

nice

Quoted 1 line

inside the header for programmers who are looking at using these

functions.

As I wrote in another email I sent yesterday, I agree with you.

In the next version, highmem.rst file will only have a little introduction about why we need to differentiate High Memory (Highmem) and the permanently mapped Low Memory (Normal), a brief description of the set functions we have to map Highmem, why we have them and which of them developers should avoid to use in new code.

Most of the information, especially how to use them, will be moved to highmem.h and highmem-internal.h.

Of course highmem.rst will automatically include this information as the kdoc directives are processed.

Quoted 4 lines

Here is an example of how I dealt with this on a recent auxiliary bus documentation update:

https://lore.kernel.org/lkml/[email protected]/

Thanks for this link. I have already skimmed through it. I think I understand the overall design. I'll rework this series taking into account what I read in your patch.

Quoted 3 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 75

Quoted 3 lines
Documentation/vm/highmem.rst +0 -0
 * kmap_atomic().  This permits a very short duration mapping of a 

single

Quoted 2 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 77

Quoted 2 lines
Documentation/vm/highmem.rst +0 -0
   page.  Since the mapping is restricted to the CPU that issued it, it
   performs well, but the issuing task is therefore required to stay on 

that

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 78

Quoted 1 line
Documentation/vm/highmem.rst +0 -0
   CPU until it has finished, lest some other task displace its 

mappings.

Quoted 2 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 67

Quoted 2 lines
Documentation/vm/highmem.rst +0 -1
-  kmap_atomic() may also be used by interrupt contexts, since it is 

does not

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 68

Quoted 1 line
Documentation/vm/highmem.rst +0 -1
-  sleep and the caller may not sleep until after kunmap_atomic() is 

called.

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 79

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  kmap_atomic() may also be used by interrupt contexts, since it does 

not

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 80

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  sleep and the caller too can not sleep until after kunmap_atomic() 

is called.

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 81

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  Each call of kmap_atomic() in the kernel creates a non-preemptible 

section and

Quoted 3 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings

Quoted 3 lines
Documentation/vm/highmem.rst +3 -0
+  disable pagefaults.
+
+  This could be a source of unwanted latency, so it should be only 

used if it is

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 85

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  absolutely required, otherwise the corresponding kmap_local_*() 

variant should

Quoted 3 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings

Quoted 3 lines
Documentation/vm/highmem.rst +3 -0
+  be used if it is feasible.
+
+  On 64-bit systems, calls to kmap() and kmap_atomic() have no real 

work to do

Quoted 2 lines
                                 ^^^^^^^^^^^^^^^^^^^^^^^^

What about kmap_local_page()?

Yes, correct. I overlooked it.

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 89

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  because a 64-bit address space is more than sufficient to address 

all the

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 90

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  physical memory, so all of physical memory appears in the direct 

mapping.

Quoted 5 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings

Quoted 5 lines
Documentation/vm/highmem.rst +3 -1
+
+  It is assumed that k[un]map_atomic() won't fail.
 
-  It may be assumed that k[un]map_atomic() won't fail.
+* kmap().  This should be used to make short duration mapping of a 

single

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 95

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  page with no restrictions on preemption or migration. It comes with 

an

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 96

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  overhead as mapping space is restricted and protected by a global 

lock

Quoted 1 line

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 97

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  for synchronization. When mapping is no more needed, the address 

that

Quoted 4 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings

Quoted 4 lines
Documentation/vm/highmem.rst +4 -0
+  the page was mapped to must be released with kunmap().
+
+  Mapping changes must be propagated across all the CPUs. kmap() also
+  requires global TLB invalidation when the kmap's pool wraps and it 

might

Quoted 2 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 103

Quoted 2 lines
Documentation/vm/highmem.rst +2 -0
+  block when the mapping space is fully utilized until a slot becomes
+  available. Therefore, kmap() is only callable from preemptible 

context.

Quoted 2 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 105

Quoted 2 lines
Documentation/vm/highmem.rst +2 -0
+
+  All the above work is necessary if a mapping must last for a 

relatively

Quoted 3 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings

Quoted 3 lines
Documentation/vm/highmem.rst +3 -0
+  long time but the bulk of high-memory mappings in the kernel are
+  short-lived and only used in one place. This means that the cost of
+  kmap() is mostly wasted in such cases. kmap() was not intended for 

long

Quoted 2 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 110

Quoted 2 lines
Documentation/vm/highmem.rst +2 -0
+  term mappings but it has morphed in that direction and its use is
+  strongly discouraged in newer code and the set of the preceding 

functions

Quoted 3 lines

Documentation/vm/highmem.rst Temporary Virtual Mappings · line 113

Quoted 1 line
Documentation/vm/highmem.rst +1 -0
+  should be preferred.

I like this paragraph especially the last sentence! Nicely put!

Thanks!

Quoted 2 lines

But this is another reason I think this text in the code next to the kmap calls. Then it is more readily available to programmers who are looking

at the

Quoted 1 line

functions.

As I wrote above, I will be reworking this series with your suggestions in mind, but probably (due to other commitments) I will not be able to submit the next version before the weekend.

Quoted 2 lines

Thanks for the great work so far!
Ira

Thanks to you for this review and all the suggestions you have provided me,

Fabio

Fabio

#4 lore raw

PATCH v3 2/2 Documentation/vm: Include kernel-doc to highmem.rst

In reply to Fabio M. De Francesco

Include kernel-doc comments to highmem.rst from include/kernel/highmem.h.

From now on, the "Using kmap_atomic" section is redundant and obsolete, therefore remove it.

Cc Jonathan Corbet <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Ira Weiny <[email protected]>
Cc Matthew Wilcox <[email protected]>
Cc Peter Zijlstra <[email protected]>
---
 Documentation/vm/highmem.rst | 41 ++++++------------------------------
 1 file changed, 6 insertions(+), 35 deletions(-)
Documentation/vm/highmem.rst +6 -35
index 12dcfbee094d..5dcee6233cd5 100644
@@ -117,41 +117,6 @@ The kernel contains several ways of creating temporary mappings:
   synchronization to unmap.
 
 
-Using kmap_atomic
-=================
-
-When and where to use kmap_atomic() is straightforward.  It is used when code
-wants to access the contents of a page that might be allocated from high memory
-(see __GFP_HIGHMEM), for example a page in the pagecache.  The API has two
-functions, and they can be used in a manner similar to the following::
-
-	/* Find the page of interest. */
-	struct page *page = find_get_page(mapping, offset);
-
-	/* Gain access to the contents of that page. */
-	void *vaddr = kmap_atomic(page);
-
-	/* Do something to the contents of that page. */
-	memset(vaddr, 0, PAGE_SIZE);
-
-	/* Unmap that page. */
-	kunmap_atomic(vaddr);
-
-Note that the kunmap_atomic() call takes the result of the kmap_atomic() call
-not the argument.
-
-If you need to map two pages because you want to copy from one page to
-another you need to keep the kmap_atomic calls strictly nested, like::
-
-	vaddr1 = kmap_atomic(page1);
-	vaddr2 = kmap_atomic(page2);
-
-	memcpy(vaddr1, vaddr2, PAGE_SIZE);
-
-	kunmap_atomic(vaddr2);
-	kunmap_atomic(vaddr1);
-
-
 Cost of Temporary Mappings
 ==========================
 
@@ -190,3 +155,9 @@ The general recommendation is that you don't use more than 8GiB on a 32-bit
 machine - although more might work for you and your workload, you're pretty
 much on your own - don't expect kernel developers to really care much if things
 come apart.
+
+
+Functions
+=========
+
+.. kernel-doc:: include/linux/highmem.h
Signature · 1 line
2.34.1
Reviewed-by on v3 2/2
#5 lore raw

Re: [PATCH v3 2/2] Documentation/vm: Include kernel-doc to highmem.rst

Quoted 6 lines · Fabio M. De Francesco

Include kernel-doc comments to highmem.rst from include/kernel/highmem.h.

From now on, the "Using kmap_atomic" section is redundant and obsolete, therefore remove it.

Quoted 6 lines
Cc Jonathan Corbet <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc Ira Weiny <[email protected]>
Cc Matthew Wilcox <[email protected]>
Cc Peter Zijlstra <[email protected]>
Quoted 63 lines
---
 Documentation/vm/highmem.rst | 41 ++++++------------------------------
 1 file changed, 6 insertions(+), 35 deletions(-)
Documentation/vm/highmem.rst +6 -35
index 12dcfbee094d..5dcee6233cd5 100644
@@ -117,41 +117,6 @@ The kernel contains several ways of creating temporary mappings:
   synchronization to unmap.
 
 
-Using kmap_atomic
-=================
-
-When and where to use kmap_atomic() is straightforward.  It is used when code
-wants to access the contents of a page that might be allocated from high memory
-(see __GFP_HIGHMEM), for example a page in the pagecache.  The API has two
-functions, and they can be used in a manner similar to the following::
-
-	/* Find the page of interest. */
-	struct page *page = find_get_page(mapping, offset);
-
-	/* Gain access to the contents of that page. */
-	void *vaddr = kmap_atomic(page);
-
-	/* Do something to the contents of that page. */
-	memset(vaddr, 0, PAGE_SIZE);
-
-	/* Unmap that page. */
-	kunmap_atomic(vaddr);
-
-Note that the kunmap_atomic() call takes the result of the kmap_atomic() call
-not the argument.
-
-If you need to map two pages because you want to copy from one page to
-another you need to keep the kmap_atomic calls strictly nested, like::
-
-	vaddr1 = kmap_atomic(page1);
-	vaddr2 = kmap_atomic(page2);
-
-	memcpy(vaddr1, vaddr2, PAGE_SIZE);
-
-	kunmap_atomic(vaddr2);
-	kunmap_atomic(vaddr1);
-
-
 Cost of Temporary Mappings
 ==========================
 
@@ -190,3 +155,9 @@ The general recommendation is that you don't use more than 8GiB on a 32-bit
 machine - although more might work for you and your workload, you're pretty
 much on your own - don't expect kernel developers to really care much if things
 come apart.
+
+
+Functions
+=========
+
+.. kernel-doc:: include/linux/highmem.h
Signature · 1 line
2.34.1
#6 lore raw
In reply to Fabio M. De Francesco
Quoted 1 line · On sabato 16 aprile 2022 01:19:15 CEST Fabio M. De Francesco wrote:

Extend and rework the "Temporary Virtual Mappings" section of the

highmem.rst

Quoted 3 lines

documentation.

Do a partial rework of the paragraph related to kmap() and add a new

paragraph

Quoted 1 line

in order to document the set of kmap_local_*() functions. Re-order

paragraphs

Quoted 7 lines

in decreasing order of preference of usage.

Include kerneldoc comments from include/linux/highmem.h and remove redundant and obsolete section about kmap_atomic().

v1-v2:
1/2 - According to comments on v1 by Matthew Wilcox and Ira

Weiny,

Quoted 10 lines

correct a mistake in text, format paragraphs to stay within the 75 characters limit, re-order the flow of the same paragraphs in decreasing order of preference of usage.
2/2 - No patch.

v2->v3:
1/2 - No changes.
2/2 - Added patch to include kernel-doc to highmem.rst and
remove the now redundant section about kmap_atomic().

Quoted 11 lines

Fabio M. De Francesco (2):
Documentation/vm: Extend "Temporary Virtual Mappings" section in
highmem.rst
Documentation/vm: Include kernel-doc to highmem.rst

 Documentation/vm/highmem.rst | 96 +++++++++++++++++++++---------------
 1 file changed, 56 insertions(+), 40 deletions(-)
Signature · 1 line
2.34.1

Dear Maintainers,

Please drop this patch because it has been reworked and inserted as part of a new series whose subject is "Extend and reorganize Highmem's documentation" and which is about to be submitted.

Thanks,

Fabio M. De Francesco