With the commit below, fat now correctly handles a file read under a non-cluster-aligned root directory of fat12/16. Write operation should be fixed in the same manner.

Fixes: commit 9b18358dc05d ("fs: fat: fix reading non-cluster-aligned root directory") Signed-off-by: AKASHI Takahiro takahiro.akashi@linaro.org Cc: Anssi Hannula anssi.hannula@bitwise.fi Tested-by: Heinrich Schuchardt xypron.glpk@gmx.de --- fs/fat/fat_write.c | 78 +++++++++++++++++++++++++++++++--------------- 1 file changed, 53 insertions(+), 25 deletions(-)

diff --git a/fs/fat/fat_write.c b/fs/fat/fat_write.c index 852f874e5817..3bc0dd637521 100644 --- a/fs/fat/fat_write.c +++ b/fs/fat/fat_write.c @@ -388,29 +388,23 @@ static __u32 determine_fatent(fsdata *mydata, __u32 entry) }

/** - * set_cluster() - write data to cluster + * set_sectors() - write data to sectors * - * Write 'size' bytes from 'buffer' into the specified cluster. + * Write 'size' bytes from 'buffer' into the specified sector. * * @mydata: data to be written - * @clustnum: cluster to be written to + * @startsect: sector to be written to * @buffer: data to be written * @size: bytes to be written (but not more than the size of a cluster) * Return: 0 on success, -1 otherwise */ static int -set_cluster(fsdata *mydata, u32 clustnum, u8 *buffer, u32 size) +set_sectors(fsdata *mydata, u32 startsect, u8 *buffer, u32 size) { - u32 idx = 0; - u32 startsect; + u32 nsects = 0; int ret;

- if (clustnum > 0) - startsect = clust_to_sect(mydata, clustnum); - else - startsect = mydata->rootdir_sect; - - debug("clustnum: %d, startsect: %d\n", clustnum, startsect); + debug("startsect: %d\n", startsect);

if ((unsigned long)buffer & (ARCH_DMA_MINALIGN - 1)) { ALLOC_CACHE_ALIGN_BUFFER(__u8, tmpbuf, mydata->sect_size); @@ -429,17 +423,16 @@ set_cluster(fsdata *mydata, u32 clustnum, u8 *buffer, u32 size) size -= mydata->sect_size; } } else if (size >= mydata->sect_size) { - idx = size / mydata->sect_size; - ret = disk_write(startsect, idx, buffer); - if (ret != idx) { + nsects = size / mydata->sect_size; + ret = disk_write(startsect, nsects, buffer); + if (ret != nsects) { debug("Error writing data (got %d)\n", ret); return -1; }

- startsect += idx; - idx *= mydata->sect_size; - buffer += idx; - size -= idx; + startsect += nsects; + buffer += nsects * mydata->sect_size; + size -= nsects * mydata->sect_size; }

if (size) { @@ -457,6 +450,44 @@ set_cluster(fsdata *mydata, u32 clustnum, u8 *buffer, u32 size) return 0; }

+/** + * set_cluster() - write data to cluster + * + * Write 'size' bytes from 'buffer' into the specified cluster. + * + * @mydata: data to be written + * @clustnum: cluster to be written to + * @buffer: data to be written + * @size: bytes to be written (but not more than the size of a cluster) + * Return: 0 on success, -1 otherwise + */ +static int +set_cluster(fsdata *mydata, u32 clustnum, u8 *buffer, u32 size) +{ + return set_sectors(mydata, clust_to_sect(mydata, clustnum), + buffer, size); +} + +static int +flush_dir(fat_itr *itr) +{ + fsdata *mydata = itr->fsdata; + u32 startsect, sect_offset, nsects; + + if (!itr->is_root || mydata->fatsize == 32) + return set_cluster(mydata, itr->clust, itr->block, + mydata->clust_size * mydata->sect_size); + + sect_offset = itr->clust * mydata->clust_size; + startsect = mydata->rootdir_sect + sect_offset; + /* do not write past the end of rootdir */ + nsects = min_t(u32, mydata->clust_size, + mydata->rootdir_size - sect_offset); + + return set_sectors(mydata, startsect, itr->block, + nsects * mydata->sect_size); +} + static __u8 tmpbuf_cluster[MAX_CLUSTSIZE] __aligned(ARCH_DMA_MINALIGN);

/* @@ -1171,8 +1202,7 @@ int file_fat_write_at(const char *filename, loff_t pos, void *buffer, }

/* Write directory table to device */ - ret = set_cluster(mydata, itr->clust, itr->block, - mydata->clust_size * mydata->sect_size); + ret = flush_dir(itr); if (ret) { printf("Error: writing directory entry\n"); ret = -EIO; @@ -1249,8 +1279,7 @@ static int delete_dentry(fat_itr *itr) memset(dentptr, 0, sizeof(*dentptr)); dentptr->name[0] = 0xe5;

- if (set_cluster(mydata, itr->clust, itr->block, - mydata->clust_size * mydata->sect_size) != 0) { + if (flush_dir(itr)) { printf("error: writing directory entry\n"); return -EIO; } @@ -1452,8 +1481,7 @@ int fat_mkdir(const char *new_dirname) }

/* Write directory table to device */ - ret = set_cluster(mydata, itr->clust, itr->block, - mydata->clust_size * mydata->sect_size); + ret = flush_dir(itr); if (ret) printf("Error: writing directory entry\n");

When a long name directory entry is created, multiple directory entries may be occupied across a directory cluster boundary. Since only one directory cluster is cached in a directory iterator, a first cluster must be written back to device before switching over a second cluster.

Without this patch, some added files may be lost even if you don't see any failures on write operation.

Signed-off-by: AKASHI Takahiro takahiro.akashi@linaro.org Tested-by: Heinrich Schuchardt xypron.glpk@gmx.de --- fs/fat/fat_write.c | 33 ++++++++++++++------------------- 1 file changed, 14 insertions(+), 19 deletions(-)

diff --git a/fs/fat/fat_write.c b/fs/fat/fat_write.c index 3bc0dd637521..da1753d545ac 100644 --- a/fs/fat/fat_write.c +++ b/fs/fat/fat_write.c @@ -209,7 +209,8 @@ name11_12: return 1; }

-static int flush_dir_table(fat_itr *itr); +static int new_dir_table(fat_itr *itr); +static int flush_dir(fat_itr *itr);

/* * Fill dir_slot entries with appropriate name, id, and attr @@ -242,19 +243,15 @@ fill_dir_slot(fat_itr *itr, const char *l_name) memcpy(itr->dent, slotptr, sizeof(dir_slot)); slotptr--; counter--; + + if (itr->remaining == 0) + flush_dir(itr); + if (!fat_itr_next(itr)) - if (!itr->dent && !itr->is_root && flush_dir_table(itr)) + if (!itr->dent && !itr->is_root && new_dir_table(itr)) return -1; }

- if (!itr->dent && !itr->is_root) - /* - * don't care return value here because we have already - * finished completing an entry with name, only ending up - * no more entry left - */ - flush_dir_table(itr); - return 0; }

@@ -621,18 +618,14 @@ static int find_empty_cluster(fsdata *mydata) }

/* - * Write directory entries in itr's buffer to block device + * Allocate a cluster for additional directory entries */ -static int flush_dir_table(fat_itr *itr) +static int new_dir_table(fat_itr *itr) { fsdata *mydata = itr->fsdata; int dir_newclust = 0; unsigned int bytesperclust = mydata->clust_size * mydata->sect_size;

- if (set_cluster(mydata, itr->clust, itr->block, bytesperclust) != 0) { - printf("error: writing directory entry\n"); - return -1; - } dir_newclust = find_empty_cluster(mydata); set_fatent_value(mydata, itr->clust, dir_newclust); if (mydata->fatsize == 32) @@ -987,7 +980,7 @@ static dir_entry *find_directory_entry(fat_itr *itr, char *filename) return itr->dent; }

- if (!itr->dent && !itr->is_root && flush_dir_table(itr)) + if (!itr->dent && !itr->is_root && new_dir_table(itr)) /* indicate that allocating dent failed */ itr->dent = NULL;

@@ -1172,14 +1165,16 @@ int file_fat_write_at(const char *filename, loff_t pos, void *buffer,

memset(itr->dent, 0, sizeof(*itr->dent));

- /* Set short name to set alias checksum field in dir_slot */ + /* Calculate checksum for short name */ set_name(itr->dent, filename); + + /* Set long name entries */ if (fill_dir_slot(itr, filename)) { ret = -EIO; goto exit; }

- /* Set attribute as archive for regular file */ + /* Set short name entry */ fill_dentry(itr->fsdata, itr->dent, filename, 0, size, 0x20);

retdent = itr->dent;

Contrary to fat12/16, fat32 can have root directory at any location and its size can be expanded. Without this patch, root directory won't grow properly and so we will eventually fail to add files under root directory. Please note that this can happen even if you delete many files as deleted directory entries are not reclaimed but just marked as "deleted" under the current implementation.

Signed-off-by: AKASHI Takahiro takahiro.akashi@linaro.org Tested-by: Heinrich Schuchardt xypron.glpk@gmx.de --- fs/fat/fat_write.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/fs/fat/fat_write.c b/fs/fat/fat_write.c index da1753d545ac..3bb9eac097eb 100644 --- a/fs/fat/fat_write.c +++ b/fs/fat/fat_write.c @@ -247,8 +247,11 @@ fill_dir_slot(fat_itr *itr, const char *l_name) if (itr->remaining == 0) flush_dir(itr);

+ /* allocate a cluster for more entries */ if (!fat_itr_next(itr)) - if (!itr->dent && !itr->is_root && new_dir_table(itr)) + if (!itr->dent && + (!itr->is_root || itr->fsdata->fatsize == 32) && + new_dir_table(itr)) return -1; }

@@ -980,7 +983,10 @@ static dir_entry *find_directory_entry(fat_itr *itr, char *filename) return itr->dent; }

- if (!itr->dent && !itr->is_root && new_dir_table(itr)) + /* allocate a cluster for more entries */ + if (!itr->dent && + (!itr->is_root || itr->fsdata->fatsize == 32) && + new_dir_table(itr)) /* indicate that allocating dent failed */ itr->dent = NULL;

Two test cases are added under test_fs_ext: test case 10: for root directory test case 11: for non-root directory

Those will verify a behavior fixed by the commits related to root directory ("fs: fat: allocate a new cluster for root directory of fat32" and "fs: fat: flush a directory cluster properly"), and focus on handling long-file-name directory entries under a directory.

Signed-off-by: AKASHI Takahiro takahiro.akashi@linaro.org --- test/py/tests/test_fs/test_ext.py | 84 +++++++++++++++++++++++++++++++ 1 file changed, 84 insertions(+)

diff --git a/test/py/tests/test_fs/test_ext.py b/test/py/tests/test_fs/test_ext.py index 2c47738b8df2..361f440dd6d4 100644 --- a/test/py/tests/test_fs/test_ext.py +++ b/test/py/tests/test_fs/test_ext.py @@ -233,3 +233,87 @@ class TestFsExt(object): % (fs_type, ADDR, MIN_FILE)]) assert('Unable to write "/dir1' in ''.join(output)) assert_fs_integrity(fs_type, fs_img) + + def test_fs_ext10(self, u_boot_console, fs_obj_ext): + """ + 'Test Case 10 - create/delete as many directories under root directory + as amount of directory entries goes beyond one cluster size)' + """ + fs_type,fs_img,md5val = fs_obj_ext + with u_boot_console.log.section('Test Case 10 - create/delete (many)'): + # Test Case 10a - Create many files + # Please note that the size of directory entry is 32 bytes. + # So one typical cluster may holds 64 (2048/32) entries. + output = u_boot_console.run_command( + 'host bind 0 %s' % fs_img) + + for i in range(0, 66): + output = u_boot_console.run_command( + '%swrite host 0:0 %x /FILE0123456789_%02x 100' + % (fs_type, ADDR, i)) + output = u_boot_console.run_command('%sls host 0:0 /' % fs_type) + assert('FILE0123456789_00' in output) + assert('FILE0123456789_41' in output) + + # Test Case 10b - Delete many files + for i in range(0, 66): + output = u_boot_console.run_command( + '%srm host 0:0 /FILE0123456789_%02x' + % (fs_type, i)) + output = u_boot_console.run_command('%sls host 0:0 /' % fs_type) + assert(not 'FILE0123456789_00' in output) + assert(not 'FILE0123456789_41' in output) + + # Test Case 10c - Create many files again + # Please note no.64 and 65 are intentionally re-created + for i in range(64, 128): + output = u_boot_console.run_command( + '%swrite host 0:0 %x /FILE0123456789_%02x 100' + % (fs_type, ADDR, i)) + output = u_boot_console.run_command('%sls host 0:0 /' % fs_type) + assert('FILE0123456789_40' in output) + assert('FILE0123456789_79' in output) + + assert_fs_integrity(fs_type, fs_img) + + def test_fs_ext11(self, u_boot_console, fs_obj_ext): + """ + 'Test Case 11 - create/delete as many directories under non-root + directory as amount of directory entries goes beyond one cluster size)' + """ + fs_type,fs_img,md5val = fs_obj_ext + with u_boot_console.log.section('Test Case 10 - create/delete (many)'): + # Test Case 11a - Create many files + # Please note that the size of directory entry is 32 bytes. + # So one typical cluster may holds 64 (2048/32) entries. + output = u_boot_console.run_command( + 'host bind 0 %s' % fs_img) + + for i in range(0, 66): + output = u_boot_console.run_command( + '%swrite host 0:0 %x /dir1/FILE0123456789_%02x 100' + % (fs_type, ADDR, i)) + output = u_boot_console.run_command('%sls host 0:0 /dir1' % fs_type) + assert('FILE0123456789_00' in output) + assert('FILE0123456789_41' in output) + + # Test Case 11b - Delete many files + for i in range(0, 66): + output = u_boot_console.run_command( + '%srm host 0:0 /dir1/FILE0123456789_%02x' + % (fs_type, i)) + output = u_boot_console.run_command('%sls host 0:0 /dir1' % fs_type) + assert(not 'FILE0123456789_00' in output) + assert(not 'FILE0123456789_41' in output) + + # Test Case 11c - Create many files again + # Please note no.64 and 65 are intentionally re-created + for i in range(64, 128): + output = u_boot_console.run_command( + '%swrite host 0:0 %x /dir1/FILE0123456789_%02x 100' + % (fs_type, ADDR, i)) + output = u_boot_console.run_command('%sls host 0:0 /dir1' % fs_type) + assert('FILE0123456789_40' in output) + assert('FILE0123456789_79' in output) + + assert_fs_integrity(fs_type, fs_img)