#
321459 |
|
25-Jul-2017 |
alc |
MFC r320077 Change blist_alloc()'s allocation policy from first-fit to next-fit so that disk writes are more likely to be sequential. This change is beneficial on both the solid state and mechanical disks that I've tested. (A similar change in allocation policy was made by DragonFly BSD in 2013 to speed up Poudriere with "stressful memory parameters".)
Increase the width of blst_meta_alloc()'s parameter "skip" and the local variables whose values are derived from it to 64 bits. (This matches the width of the field "skip" that is stored in the structure "blist" and passed to blst_meta_alloc().)
Eliminate a pointless check for a NULL blist_t.
Simplify blst_meta_alloc()'s handling of the ALL-FREE case.
Address nearby style errors.
MFC r320417 Address the remaining integer overflow issues with the "skip" parameters and "next_skip" variables. The "skip" value in struct blist has long been a 64-bit quantity but various functions have implicitly truncated this value to 32 bits. Now, all arithmetic involving the "skip" value is 64 bits wide. (This should allow us to relax the size limit on a swap device in the swap pager.)
Maintain the ability to test this allocator as a user-space application by including <stdbool.h>.
Remove an unused variable from blst_radix_print().
MFC r320527 Change blst_leaf_alloc() to handle a cursor argument, and to improve performance.
To find in the leaf bitmap all ranges of sufficient length, use a doubling strategy with shift-and-and until each bit still set represents a bit sequence of length 'count', or until the bitmask is zero. In the latter case, update the hint based on the first bit sequence length not found to be available. For example, seeking an interval of length 12, the set bits of the bitmap would represent intervals of length 1, then 2, then 3, then 6, then 12. If no bits are set at the point when each bit represents an interval of length 6, then the hint can be updated to 5 and the search terminated.
If long-enough intervals are found, discard those before the cursor. If any remain, use binary search to find the position of the first of them, and allocate that interval.
|
#
321375 |
|
22-Jul-2017 |
alc |
MFC r319905
Reduce the frequency of hint updates on allocation without incurring additional allocation overhead. Previously, blst_meta_alloc() updated the hint after every successful allocation. However, these "eager" hint updates are of no actual benefit if, instead, the "lazy" hint update at the start of blst_meta_alloc() is generalized to handle all cases where the number of available blocks is less than the requested allocation. Previously, the lazy hint update at the start of blst_meta_alloc() only handled the ALL-FULL case. (I would also note that this change provides consistency between blist_alloc() and blist_fill() in that their hint maintenance is now entirely lazy.)
Eliminate unnecessary checks for terminators in blst_meta_alloc() and blst_meta_fill() when handling ALL-FREE meta nodes.
Eliminate the field "bl_free" from struct blist. It is redundant. Unless the entire radix tree is a single leaf, the count of free blocks is stored in the root node. Instead, provide a function blist_avail() for obtaining the number of free blocks.
In blst_meta_alloc(), perform a sanity check on the allocation once rather than repeating it in a loop over the meta node's children.
In blst_leaf_fill(), use the optimized bitcount*() function instead of a loop to count the blocks being allocated.
Add or improve several comments.
Address some nearby style errors.
|
#
321365 |
|
22-Jul-2017 |
alc |
MFC r319756 Style and comment fixes
|
#
320622 |
|
03-Jul-2017 |
alc |
MFC r319699 When allocating swap blocks, if the available number of free blocks in a subtree is already zero, then setting the "largest contiguous free block" hint for that subtree to anything other than zero makes no sense. (To be clear, assigning a value to the hint that is too large is not a correctness problem, only a pessimization.)
MFC r319755 blist_fill()'s return type is too narrow. blist_fill() accepts a 64-bit quantity as the size of the range to fill, but returns a 32-bit quantity as the number of blocks that were allocated to fill that range. This revision corrects that mismatch.
MFC r319793 Remove an unnecessary field from struct blist. (The comment describing what this field represented was also inaccurate.)
In r178792, blist_create() grew a malloc flag, allowing M_NOWAIT to be specified. However, blist_create() was not modified to handle the possibility that a malloc() call failed. Address this omission.
Increase the width of the local variable "radix" to 64 bits. This matches the width of the corresponding field in struct blist.
|
#
319965 |
|
15-Jun-2017 |
alc |
MFC r318995 In r118390, the swap pager's approach to striping swap allocation over multiple devices was changed. However, swapoff_one() was not fully and correctly converted. In particular, with r118390's introduction of a per- device blist, the maximum swap block size, "dmmax", became irrelevant to swapoff_one()'s operation. Moreover, swapoff_one() was performing out-of- range operations on the per-device blist that were silently ignored by blist_fill().
This change corrects both of these problems with swapoff_one(), which will allow us to potentially increase MAX_PAGEOUT_CLUSTER. Previously, swapoff_one() would panic inside of blist_fill() if you increased MAX_PAGEOUT_CLUSTER.
MFC r319001 After r118390, the variable "dmmax" was neither the correct strip size nor the correct maximum block size. Moreover, after r318995, it serves no purpose except to provide information to user space through a read- sysctl.
This change eliminates the variable "dmmax" but retains the sysctl. It also corrects the value returned by the sysctl.
MFC r319604 Halve the memory being internally allocated by the blist allocator. In short, half of the memory that is allocated to implement the radix tree is wasted because we did not change "u_daddr_t" to be a 64-bit unsigned int when we changed "daddr_t" to be a 64-bit (signed) int. (See r96849 and r96851.)
MFC r319612 When the function blist_fill() was added to the kernel in r107913, the swap pager used a different scheme for striping the allocation of swap space across multiple devices. And, although blist_fill() was intended to support fill operations with large counts, the old striping scheme never performed a fill larger than the stripe size. Consequently, the misplacement of a sanity check in blst_meta_fill() went undetected. Now, moving forward in time to r118390, a new scheme for striping was introduced that maintained a blist allocator per device, but as noted in r318995, swapoff_one() was not fully and correctly converted to the new scheme. This change completes what was started in r318995 by fixing the underlying bug in blst_meta_fill() that stops swapoff_one() from simply performing a single blist_fill() operation.
MFC r319627 Starting in r118390, swaponsomething() began to reserve the blocks at the beginning of a swap area for a disk label. However, neither r118390 nor r118544, which increased the reservation from one to two blocks, correctly accounted for these blocks when updating the variable "swap_pager_avail". This change corrects that error.
MFC r319655 Originally, this file could be compiled as a user-space application for testing purposes. However, over the years, various changes to the kernel have broken this feature. This revision applies some fixes to get user- space compilation working again. There are no changes in this revision to code that is used by the kernel.
|
#
256281 |
|
10-Oct-2013 |
gjb |
Copy head (r256279) to stable/10 as part of the 10.0-RELEASE cycle.
Approved by: re (implicit) Sponsored by: The FreeBSD Foundation |
#
178792 |
|
05-May-2008 |
kmacy |
add malloc flag to blist so that it can be used in ithread context
Reviewed by: alc, bsdimp
|
#
139825 |
|
07-Jan-2005 |
imp |
/* -> /*- for license, minor formatting changes
|
#
130049 |
|
04-Jun-2004 |
alc |
Move the definitions of SWAPBLK_NONE and SWAPBLK_MASK from vm_page.h to blist.h, enabling the removal of numerous #includes from subr_blist.c. (subr_blist.c and swap_pager.c are the only users of these definitions.)
|
#
118848 |
|
12-Aug-2003 |
imp |
Expand inline the relevant parts of src/COPYRIGHT for Matt Dillon's copyrighted files.
Approved by: Matt Dillon
|
#
109086 |
|
10-Jan-2003 |
dillon |
Remove all use of the LOG2() macro/inline, undoing some non-optimal cruft that crept in recently. GCC will optimize the divides and multiplies for us.
Submitted by: David Schultz <dschultz@uclink.Berkeley.EDU> MFC after: 1 day
|
#
107913 |
|
15-Dec-2002 |
dillon |
This is David Schultz's swapoff code which I am finally able to commit. This should be considered highly experimental for the moment.
Submitted by: David Schultz <dschultz@uclink.Berkeley.EDU> MFC after: 3 weeks
|
#
96849 |
|
18-May-2002 |
phk |
Move the hideously misnamed type "u_daddr_t" to <sys/blist.h> where it belongs.
Sponsored by: DARPA & NAI Labs.
|
#
96567 |
|
14-May-2002 |
phk |
Roll the LOG2 macro up again, I don't belive unrolling this for 64bits make sense.
Sponsored by: DARPA & NAI Labs.
|
#
50477 |
|
27-Aug-1999 |
peter |
$Id$ -> $FreeBSD$
|
#
42956 |
|
21-Jan-1999 |
dillon |
Add new blist module - radix tree based bitmap allocator with size hinting. Will be used by the new swapper.
|