From c001596110e834a85b01a47a20811b318cb3b9e4 Mon Sep 17 00:00:00 2001 From: Theodore Ts'o Date: Fri, 26 Feb 2021 17:41:06 -0500 Subject: [PATCH] libext2fs: fix unix_io's Direct I/O support The previous Direct I/O support worked on HDD's with 512 byte logical sector sizes, and on FreeBSD which required 4k aligned memory buffers. However, it was incomplete and was not correctly working on HDD's with 4k logical sector sizes (aka Advanced Format Disks). Based on a patch from Alexey Lyashkov but rewritten to work with the latest e2fsprogs and to minimize changes to make it easier to review. Signed-off-by: Theodore Ts'o Reported-by: Artem Blagodarenko Signed-off-by: Pavel Reichl --- lib/ext2fs/io_manager.c | 6 ++- lib/ext2fs/unix_io.c | 87 +++++++++++++++++++++++++++++++---------- 2 files changed, 70 insertions(+), 23 deletions(-) Index: e2fsprogs-1.45.6/lib/ext2fs/io_manager.c =================================================================== --- e2fsprogs-1.45.6.orig/lib/ext2fs/io_manager.c +++ e2fsprogs-1.45.6/lib/ext2fs/io_manager.c @@ -134,9 +134,11 @@ errcode_t io_channel_alloc_buf(io_channe else size = -count; - if (io->align) + if (io->align) { + if (io->align > size) + size = io->align; return ext2fs_get_memalign(size, io->align, ptr); - else + } else return ext2fs_get_mem(size, ptr); } Index: e2fsprogs-1.45.6/lib/ext2fs/unix_io.c =================================================================== --- e2fsprogs-1.45.6.orig/lib/ext2fs/unix_io.c +++ e2fsprogs-1.45.6/lib/ext2fs/unix_io.c @@ -165,13 +165,15 @@ static errcode_t raw_read_blk(io_channel int actual = 0; unsigned char *buf = bufv; ssize_t really_read = 0; + unsigned long long aligned_blk; + int align_size, offset; size = (count < 0) ? -count : (ext2_loff_t) count * channel->block_size; data->io_stats.bytes_read += size; location = ((ext2_loff_t) block * channel->block_size) + data->offset; if (data->flags & IO_FLAG_FORCE_BOUNCE) { - if (ext2fs_llseek(data->dev, location, SEEK_SET) != location) { + if (ext2fs_llseek(data->dev, location, SEEK_SET) < 0) { retval = errno ? errno : EXT2_ET_LLSEEK_FAILED; goto error_out; } @@ -182,6 +184,7 @@ static errcode_t raw_read_blk(io_channel /* Try an aligned pread */ if ((channel->align == 0) || (IS_ALIGNED(buf, channel->align) && + IS_ALIGNED(location, channel->align) && IS_ALIGNED(size, channel->align))) { actual = pread64(data->dev, buf, size, location); if (actual == size) @@ -193,6 +196,7 @@ static errcode_t raw_read_blk(io_channel if ((sizeof(off_t) >= sizeof(ext2_loff_t)) && ((channel->align == 0) || (IS_ALIGNED(buf, channel->align) && + IS_ALIGNED(location, channel->align) && IS_ALIGNED(size, channel->align)))) { actual = pread(data->dev, buf, size, location); if (actual == size) @@ -201,12 +205,13 @@ static errcode_t raw_read_blk(io_channel } #endif /* HAVE_PREAD */ - if (ext2fs_llseek(data->dev, location, SEEK_SET) != location) { + if (ext2fs_llseek(data->dev, location, SEEK_SET) < 0) { retval = errno ? errno : EXT2_ET_LLSEEK_FAILED; goto error_out; } if ((channel->align == 0) || (IS_ALIGNED(buf, channel->align) && + IS_ALIGNED(location, channel->align) && IS_ALIGNED(size, channel->align))) { actual = read(data->dev, buf, size); if (actual != size) { @@ -231,21 +236,37 @@ static errcode_t raw_read_blk(io_channel * to the O_DIRECT rules, so we need to do this the hard way... */ bounce_read: + if ((channel->block_size > channel->align) && + (channel->block_size % channel->align) == 0) + align_size = channel->block_size; + else + align_size = channel->align; + aligned_blk = location / align_size; + offset = location % align_size; + + if (ext2fs_llseek(data->dev, aligned_blk * align_size, SEEK_SET) < 0) { + retval = errno ? errno : EXT2_ET_LLSEEK_FAILED; + goto error_out; + } while (size > 0) { - actual = read(data->dev, data->bounce, channel->block_size); - if (actual != channel->block_size) { + actual = read(data->dev, data->bounce, align_size); + if (actual != align_size) { actual = really_read; buf -= really_read; size += really_read; goto short_read; } - actual = size; - if (size > channel->block_size) - actual = channel->block_size; - memcpy(buf, data->bounce, actual); + if ((actual + offset) > align_size) + actual = align_size - offset; + if (actual > size) + actual = size; + memcpy(buf, data->bounce + offset, actual); + really_read += actual; size -= actual; buf += actual; + offset = 0; + aligned_blk++; } return 0; @@ -268,6 +289,8 @@ static errcode_t raw_write_blk(io_channe int actual = 0; errcode_t retval; const unsigned char *buf = bufv; + unsigned long long aligned_blk; + int align_size, offset; if (count == 1) size = channel->block_size; @@ -282,7 +305,7 @@ static errcode_t raw_write_blk(io_channe location = ((ext2_loff_t) block * channel->block_size) + data->offset; if (data->flags & IO_FLAG_FORCE_BOUNCE) { - if (ext2fs_llseek(data->dev, location, SEEK_SET) != location) { + if (ext2fs_llseek(data->dev, location, SEEK_SET) < 0) { retval = errno ? errno : EXT2_ET_LLSEEK_FAILED; goto error_out; } @@ -293,6 +316,7 @@ static errcode_t raw_write_blk(io_channe /* Try an aligned pwrite */ if ((channel->align == 0) || (IS_ALIGNED(buf, channel->align) && + IS_ALIGNED(location, channel->align) && IS_ALIGNED(size, channel->align))) { actual = pwrite64(data->dev, buf, size, location); if (actual == size) @@ -303,6 +327,7 @@ static errcode_t raw_write_blk(io_channe if ((sizeof(off_t) >= sizeof(ext2_loff_t)) && ((channel->align == 0) || (IS_ALIGNED(buf, channel->align) && + IS_ALIGNED(location, channel->align) && IS_ALIGNED(size, channel->align)))) { actual = pwrite(data->dev, buf, size, location); if (actual == size) @@ -310,13 +335,14 @@ static errcode_t raw_write_blk(io_channe } #endif /* HAVE_PWRITE */ - if (ext2fs_llseek(data->dev, location, SEEK_SET) != location) { + if (ext2fs_llseek(data->dev, location, SEEK_SET) < 0) { retval = errno ? errno : EXT2_ET_LLSEEK_FAILED; goto error_out; } if ((channel->align == 0) || (IS_ALIGNED(buf, channel->align) && + IS_ALIGNED(location, channel->align) && IS_ALIGNED(size, channel->align))) { actual = write(data->dev, buf, size); if (actual < 0) { @@ -340,37 +366,56 @@ static errcode_t raw_write_blk(io_channe * to the O_DIRECT rules, so we need to do this the hard way... */ bounce_write: + if ((channel->block_size > channel->align) && + (channel->block_size % channel->align) == 0) + align_size = channel->block_size; + else + align_size = channel->align; + aligned_blk = location / align_size; + offset = location % align_size; + while (size > 0) { - if (size < channel->block_size) { + int actual_w; + + if (size < align_size || offset) { + if (ext2fs_llseek(data->dev, aligned_blk * align_size, + SEEK_SET) < 0) { + retval = errno ? errno : EXT2_ET_LLSEEK_FAILED; + goto error_out; + } actual = read(data->dev, data->bounce, - channel->block_size); - if (actual != channel->block_size) { + align_size); + if (actual != align_size) { if (actual < 0) { retval = errno; goto error_out; } memset((char *) data->bounce + actual, 0, - channel->block_size - actual); + align_size - actual); } } actual = size; - if (size > channel->block_size) - actual = channel->block_size; - memcpy(data->bounce, buf, actual); - if (ext2fs_llseek(data->dev, location, SEEK_SET) != location) { + if ((actual + offset) > align_size) + actual = align_size - offset; + if (actual > size) + actual = size; + memcpy(((char *)data->bounce) + offset, buf, actual); + if (ext2fs_llseek(data->dev, aligned_blk * align_size, SEEK_SET) < 0) { retval = errno ? errno : EXT2_ET_LLSEEK_FAILED; goto error_out; } - actual = write(data->dev, data->bounce, channel->block_size); - if (actual < 0) { + actual_w = write(data->dev, data->bounce, align_size); + if (actual_w < 0) { retval = errno; goto error_out; } - if (actual != channel->block_size) + if (actual_w != align_size) goto short_write; size -= actual; buf += actual; location += actual; + aligned_blk++; + offset = 0; } return 0;