stasis-bLSM/datapage.cpp

407 lines
12 KiB
C++
Raw Normal View History

/*
* datapage.cpp
*
* Copyright 2009-2012 Yahoo! Inc.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*
* Author: makdere
*/
#include "logstore.h"
#include "datapage.h"
#include "regionAllocator.h"
#include <stasis/page.h>
static const int DATA_PAGE = USER_DEFINED_PAGE(1);
#define MAX_PAGE_COUNT 1000 // ~ 4MB
BEGIN_C_DECLS
static void dataPageFsck(Page* p) {
int32_t is_last_page = *stasis_page_int32_cptr_from_start(p, 0);
assert(is_last_page == 0 || is_last_page == 1 || is_last_page == 2);
}
static void dataPageLoaded(Page* p) {
dataPageFsck(p);
}
static void dataPageFlushed(Page* p) {
*stasis_page_lsn_ptr(p) = p->LSN;
dataPageFsck(p);
}
static int notSupported(int xid, Page * p) { return 0; }
END_C_DECLS
void DataPage::register_stasis_page_impl() {
static page_impl pi = {
DATA_PAGE,
1,
0, //slottedRead,
0, //slottedWrite,
0,// readDone
0,// writeDone
0,//slottedGetType,
0,//slottedSetType,
0,//slottedGetLength,
0,//slottedFirst,
0,//slottedNext,
0,//slottedLast,
notSupported, // is block supported
stasis_block_first_default_impl,
stasis_block_next_default_impl,
stasis_block_done_default_impl,
0,//slottedFreespace,
0,//slottedCompact,
0,//slottedCompactSlotIDs,
0,//slottedPreRalloc,
0,//slottedPostRalloc,
0,//slottedSpliceSlot,
0,//slottedFree,
0,//XXX page_impl_dereference_identity,
dataPageLoaded, //dataPageLoaded,
dataPageFlushed, //dataPageFlushed,
0,//slottedCleanup
};
stasis_page_impl_register(pi);
}
DataPage::DataPage(int xid, RegionAllocator * alloc, pageid_t pid): // XXX Hack!! The read-only constructor signature is too close to the other's
xid_(xid),
page_count_(1), // will be opportunistically incremented as we scan the datapage.
initial_page_count_(-1), // used by append.
alloc_(alloc), // read-only, and we don't free data pages one at a time.
first_page_(pid),
write_offset_(-1)
{
assert(pid!=0);
Page *p = alloc_ ? alloc_->load_page(xid, first_page_) : loadPage(xid, first_page_);
if(!(*is_another_page_ptr(p) == 0 || *is_another_page_ptr(p) == 2)) {
printf("Page %lld is not the start of a datapage\n", first_page_); fflush(stdout);
abort();
}
assert(*is_another_page_ptr(p) == 0 || *is_another_page_ptr(p) == 2); // would be 1 for page in the middle of a datapage
releasePage(p);
}
DataPage::DataPage(int xid, pageid_t page_count, RegionAllocator *alloc) :
xid_(xid),
page_count_(1),
initial_page_count_(page_count),
alloc_(alloc),
first_page_(alloc_->alloc_extent(xid_, page_count_)),
write_offset_(0)
{
DEBUG("Datapage page count: %lld pid = %lld\n", (long long int)initial_page_count_, (long long int)first_page_);
assert(page_count_ >= 1);
initialize();
}
void DataPage::initialize() {
initialize_page(first_page_);
}
void DataPage::initialize_page(pageid_t pageid) {
//load the first page
Page *p;
#ifdef CHECK_FOR_SCRIBBLING
p = alloc_ ? alloc->load_page(xid_, pageid) : loadPage(xid_, pageid);
if(*stasis_page_type_ptr(p) == DATA_PAGE) {
printf("Collision on page %lld\n", (long long)pageid); fflush(stdout);
assert(*stasis_page_type_ptr(p) != DATA_PAGE);
}
#else
p = loadUninitializedPage(xid_, pageid);
#endif
DEBUG("\t\t\t\t\t\t->%lld\n", pageid);
//initialize header
p->pageType = DATA_PAGE;
//clear page (arranges for null-padding) XXX null pad more carefully and use sentinel value instead?
memset(p->memAddr, 0, PAGE_SIZE);
//we're the last page for now.
*is_another_page_ptr(p) = 0;
//write 0 to first data size
*length_at_offset_ptr(p, calc_chunk_from_offset(write_offset_).slot) = 0;
//set the page dirty
stasis_page_lsn_write(xid_, p, alloc_->get_lsn(xid_));
releasePage(p);
}
size_t DataPage::write_bytes(const byte * buf, ssize_t remaining, Page ** latch_p) {
if(latch_p) { *latch_p = NULL; }
recordid chunk = calc_chunk_from_offset(write_offset_);
if(chunk.size > remaining) {
chunk.size = remaining;
}
if(chunk.page >= first_page_ + page_count_) {
chunk.size = 0; // no space (should not happen)
} else {
Page *p = alloc_ ? alloc_->load_page(xid_, chunk.page) : loadPage(xid_, chunk.page);
assert(chunk.size);
memcpy(data_at_offset_ptr(p, chunk.slot), buf, chunk.size);
stasis_page_lsn_write(xid_, p, alloc_->get_lsn(xid_));
if(latch_p && !*latch_p) {
writelock(p->rwlatch,0);
*latch_p = p;
} else {
releasePage(p);
}
write_offset_ += chunk.size;
}
return chunk.size;
}
size_t DataPage::read_bytes(byte * buf, off_t offset, ssize_t remaining) {
recordid chunk = calc_chunk_from_offset(offset);
if(chunk.size > remaining) {
chunk.size = remaining;
}
if(chunk.page >= first_page_ + page_count_) {
chunk.size = 0; // eof
} else {
Page *p = alloc_ ? alloc_->load_page(xid_, chunk.page) : loadPage(xid_, chunk.page);
if(p->pageType != DATA_PAGE) {
fprintf(stderr, "Page type %d, id %lld lsn %lld\n", (int)p->pageType, (long long)p->id, (long long)p->LSN);
assert(p->pageType == DATA_PAGE);
}
if((chunk.page + 1 == page_count_ + first_page_)
&& (*is_another_page_ptr(p))) {
page_count_++;
}
memcpy(buf, data_at_offset_ptr(p, chunk.slot), chunk.size);
releasePage(p);
}
return chunk.size;
}
bool DataPage::initialize_next_page() {
recordid rid = calc_chunk_from_offset(write_offset_);
assert(rid.slot == 0);
DEBUG("\t\t%lld\n", (long long)rid.page);
if(rid.page >= first_page_ + page_count_) {
assert(rid.page == first_page_ + page_count_);
if(alloc_->grow_extent(1)) {
page_count_++;
} else {
return false; // The region is full
}
} else {
abort();
}
Page *p = alloc_ ? alloc_->load_page(xid_, rid.page-1) : loadPage(xid_, rid.page-1);
*is_another_page_ptr(p) = (rid.page-1 == first_page_) ? 2 : 1;
stasis_page_lsn_write(xid_, p, alloc_->get_lsn(xid_));
releasePage(p);
initialize_page(rid.page);
return true;
}
Page * DataPage::write_data_and_latch(const byte * buf, size_t len, bool init_next, bool latch) {
bool first = true;
Page * p = 0;
while(1) {
assert(len > 0);
size_t written;
if(latch && first ) {
written = write_bytes(buf, len, &p);
} else {
written = write_bytes(buf, len);
}
if(written == 0) {
assert(!p);
return 0; // fail
}
if(written == len) {
if(latch) {
return p;
} else {
return (Page*)1;
}
}
if(len > PAGE_SIZE && ! first) {
assert(written > 4000);
}
buf += written;
len -= written;
if(init_next) {
if(!initialize_next_page()) {
if(p) {
unlock(p->rwlatch);
releasePage(p);
}
return 0; // fail
}
}
first = false;
}
}
bool DataPage::write_data(const byte * buf, size_t len, bool init_next) {
return 0 != write_data_and_latch(buf, len, init_next, false);
}
bool DataPage::read_data(byte * buf, off_t offset, size_t len) {
while(1) {
assert(len > 0);
size_t read_count = read_bytes(buf, offset, len);
if(read_count == 0) {
return false; // fail
}
if(read_count == len) {
return true; // success
}
buf += read_count;
offset += read_count;
len -= read_count;
}
}
bool DataPage::append(datatuple const * dat)
{
// First, decide if we should append to this datapage, based on whether
// appending will waste more or less space than starting a new datapage
bool accept_tuple;
len_t tup_len = dat->byte_length();
// Decsion tree
if(write_offset_ > (initial_page_count_ * PAGE_SIZE)) {
// we already exceeded the page budget
if(write_offset_ > (2 * initial_page_count_ * PAGE_SIZE)) {
// ... by a lot. Reject regardless. This prevents small tuples from
// being stuck behind giant ones without sacrificing much space
// (as a percentage of the whole index), because this path only
// can happen once per giant object.
accept_tuple = false;
} else {
// ... by a little bit.
accept_tuple = true;
//Accept tuple if it fits on this page, or if it's big..
//accept_tuple = (((write_offset_-1) & ~(PAGE_SIZE-1)) == (((write_offset_ + tup_len)-1) & ~(PAGE_SIZE-1)));
}
} else {
if(write_offset_ + tup_len < (initial_page_count_ * PAGE_SIZE)) {
// tuple fits. contractually obligated to accept it.
accept_tuple = true;
} else if(write_offset_ == 0) {
// datapage is empty. contractually obligated to accept tuple.
accept_tuple = true;
} else {
if(tup_len > initial_page_count_ * PAGE_SIZE) {
// this is a "big tuple"
len_t reject_padding = PAGE_SIZE - (write_offset_ & (PAGE_SIZE-1));
len_t accept_padding = PAGE_SIZE - ((write_offset_ + tup_len) & (PAGE_SIZE-1));
accept_tuple = accept_padding < reject_padding;
} else {
// this is a "small tuple"; only exceed budget if doing so leads to < 33% overhead for this data.
len_t accept_padding = PAGE_SIZE - (write_offset_ & (PAGE_SIZE-1));
accept_tuple = (3*accept_padding) < tup_len;
}
}
}
if(!accept_tuple) {
DEBUG("offset %lld closing datapage\n", write_offset_);
return false;
}
DEBUG("offset %lld continuing datapage\n", write_offset_);
// TODO could be more efficient; this does a malloc and memcpy.
// The alternative couples us more strongly to datatuple, but simplifies
// datapage.
byte * buf = dat->to_bytes();
len_t dat_len = dat->byte_length();
Page * p = write_data_and_latch((const byte*)&dat_len, sizeof(dat_len));
bool succ = false;
if(p) {
succ = write_data(buf, dat_len);
unlock(p->rwlatch);
releasePage(p);
}
free(buf);
return succ;
}
bool DataPage::recordRead(const datatuple::key_t key, size_t keySize, datatuple ** buf)
{
iterator itr(this, NULL);
int match = -1;
while((*buf=itr.getnext()) != 0) {
match = datatuple::compare((*buf)->strippedkey(), (*buf)->strippedkeylen(), key, keySize);
if(match<0) { //keep searching
datatuple::freetuple(*buf);
*buf=0;
} else if(match==0) { //found
return true;
} else { // match > 0, then does not exist
datatuple::freetuple(*buf);
*buf = 0;
break;
}
}
return false;
}
///////////////////////////////////////////////////////////////
//RECORD ITERATOR
///////////////////////////////////////////////////////////////
datatuple* DataPage::iterator::getnext() {
len_t len;
bool succ;
if(dp == NULL) { return NULL; }
// XXX hack: read latch the page that the record will live on.
// This should be handled by a read_data_in_latch function, or something...
Page * p = loadPage(dp->xid_, dp->calc_chunk_from_offset(read_offset_).page);
readlock(p->rwlatch, 0);
succ = dp->read_data((byte*)&len, read_offset_, sizeof(len));
if((!succ) || (len == 0)) {
unlock(p->rwlatch);
releasePage(p);
return NULL;
}
read_offset_ += sizeof(len);
byte * buf = (byte*)malloc(len);
succ = dp->read_data(buf, read_offset_, len);
// release hacky latch
unlock(p->rwlatch);
releasePage(p);
if(!succ) { read_offset_ -= sizeof(len); free(buf); return NULL; }
read_offset_ += len;
datatuple *ret = datatuple::from_bytes(buf);
free(buf);
return ret;
}