From 389cc9af2087aa5369ac6bf0124d14877d541966 Mon Sep 17 00:00:00 2001
From: Lennart Poettering <lennart@poettering.net>
Date: Fri, 4 Feb 2022 17:39:44 +0100
Subject: [PATCH] repart: fix sector size handling
This queries the sector size from libfdisk instead of assuming 512, and
uses that when converting from bytes to the offset/size values libfdisk
expects.
This is an alternative to Tom Yan's #21823, but prefers using libfdisk's
own ideas of the sector size instead of going directly to the backing
device via ioctls. (libfdisk can after all also operate on regular
files, where the sector size concept doesn't necessarily apply the same
way.)
This also makes the "grain" variable, i.e. how we'll align the
partitions. Previously this was hardcoded to 4K, and that still will be
the minimum grain we use, but should the sector size be larger than that
we'll use the next multiple of the sector size instead.
(cherry picked from commit 994b303123ebe6a140bf3e56c66aa66119ae7d95)
Related: #2017035
---
src/partition/repart.c | 212 +++++++++++++++++++++++++----------------
1 file changed, 132 insertions(+), 80 deletions(-)
diff --git a/src/partition/repart.c b/src/partition/repart.c
index d08f47f2c4..0862a37a8d 100644
--- a/src/partition/repart.c
+++ b/src/partition/repart.c
@@ -195,6 +195,8 @@ struct Context {
uint64_t start, end, total;
struct fdisk_context *fdisk_context;
+ uint64_t sector_size;
+ uint64_t grain_size;
sd_id128_t seed;
};
@@ -407,9 +409,12 @@ static bool context_drop_one_priority(Context *context) {
return true;
}
-static uint64_t partition_min_size(const Partition *p) {
+static uint64_t partition_min_size(Context *context, const Partition *p) {
uint64_t sz;
+ assert(context);
+ assert(p);
+
/* Calculate the disk space we really need at minimum for this partition. If the partition already
* exists the current size is what we really need. If it doesn't exist yet refuse to allocate less
* than 4K.
@@ -428,50 +433,60 @@ static uint64_t partition_min_size(const Partition *p) {
uint64_t d = 0;
if (p->encrypt != ENCRYPT_OFF)
- d += round_up_size(LUKS2_METADATA_SIZE, 4096);
+ d += round_up_size(LUKS2_METADATA_SIZE, context->grain_size);
if (p->copy_blocks_size != UINT64_MAX)
- d += round_up_size(p->copy_blocks_size, 4096);
+ d += round_up_size(p->copy_blocks_size, context->grain_size);
else if (p->format || p->encrypt != ENCRYPT_OFF) {
uint64_t f;
/* If we shall synthesize a file system, take minimal fs size into account (assumed to be 4K if not known) */
- f = p->format ? minimal_size_by_fs_name(p->format) : UINT64_MAX;
- d += f == UINT64_MAX ? 4096 : f;
+ f = p->format ? round_up_size(minimal_size_by_fs_name(p->format), context->grain_size) : UINT64_MAX;
+ d += f == UINT64_MAX ? context->grain_size : f;
}
if (d > sz)
sz = d;
}
- return MAX(p->size_min != UINT64_MAX ? p->size_min : DEFAULT_MIN_SIZE, sz);
+ return MAX(round_up_size(p->size_min != UINT64_MAX ? p->size_min : DEFAULT_MIN_SIZE, context->grain_size), sz);
}
-static uint64_t partition_max_size(const Partition *p) {
+static uint64_t partition_max_size(const Context *context, const Partition *p) {
+ uint64_t sm;
+
/* Calculate how large the partition may become at max. This is generally the configured maximum
* size, except when it already exists and is larger than that. In that case it's the existing size,
* since we never want to shrink partitions. */
+ assert(context);
+ assert(p);
+
if (PARTITION_IS_FOREIGN(p)) {
/* Don't allow changing size of partitions not managed by us */
assert(p->current_size != UINT64_MAX);
return p->current_size;
}
+ sm = round_down_size(p->size_max, context->grain_size);
+
if (p->current_size != UINT64_MAX)
- return MAX(p->current_size, p->size_max);
+ return MAX(p->current_size, sm);
- return p->size_max;
+ return sm;
}
-static uint64_t partition_min_size_with_padding(const Partition *p) {
+static uint64_t partition_min_size_with_padding(Context *context, const Partition *p) {
uint64_t sz;
/* Calculate the disk space we need for this partition plus any free space coming after it. This
* takes user configured padding into account as well as any additional whitespace needed to align
* the next partition to 4K again. */
- sz = partition_min_size(p);
+ assert(context);
+ assert(p);
+
+ sz = partition_min_size(context, p);
if (p->padding_min != UINT64_MAX)
sz += p->padding_min;
@@ -479,11 +494,11 @@ static uint64_t partition_min_size_with_padding(const Partition *p) {
if (PARTITION_EXISTS(p)) {
/* If the partition wasn't aligned, add extra space so that any we might add will be aligned */
assert(p->offset != UINT64_MAX);
- return round_up_size(p->offset + sz, 4096) - p->offset;
+ return round_up_size(p->offset + sz, context->grain_size) - p->offset;
}
/* If this is a new partition we'll place it aligned, hence we just need to round up the required size here */
- return round_up_size(sz, 4096);
+ return round_up_size(sz, context->grain_size);
}
static uint64_t free_area_available(const FreeArea *a) {
@@ -495,9 +510,12 @@ static uint64_t free_area_available(const FreeArea *a) {
return a->size - a->allocated;
}
-static uint64_t free_area_available_for_new_partitions(const FreeArea *a) {
+static uint64_t free_area_available_for_new_partitions(Context *context, const FreeArea *a) {
uint64_t avail;
+ assert(context);
+ assert(a);
+
/* Similar to free_area_available(), but takes into account that the required size and padding of the
* preceding partition is honoured. */
@@ -505,16 +523,16 @@ static uint64_t free_area_available_for_new_partitions(const FreeArea *a) {
if (a->after) {
uint64_t need, space_end, new_end;
- need = partition_min_size_with_padding(a->after);
+ need = partition_min_size_with_padding(context, a->after);
assert(a->after->offset != UINT64_MAX);
assert(a->after->current_size != UINT64_MAX);
/* Calculate where the free area ends, based on the offset of the partition preceding it */
- space_end = round_up_size(a->after->offset + a->after->current_size, 4096) + avail;
+ space_end = round_up_size(a->after->offset + a->after->current_size, context->grain_size) + avail;
/* Calculate where the partition would end when we give it as much as it needs */
- new_end = round_up_size(a->after->offset + need, 4096);
+ new_end = round_up_size(a->after->offset + need, context->grain_size);
/* Calculate saturated difference of the two: that's how much we have free for other partitions */
return LESS_BY(space_end, new_end);
@@ -523,15 +541,18 @@ static uint64_t free_area_available_for_new_partitions(const FreeArea *a) {
return avail;
}
-static int free_area_compare(FreeArea *const *a, FreeArea *const*b) {
- return CMP(free_area_available_for_new_partitions(*a),
- free_area_available_for_new_partitions(*b));
+static int free_area_compare(FreeArea *const *a, FreeArea *const*b, Context *context) {
+ assert(context);
+
+ return CMP(free_area_available_for_new_partitions(context, *a),
+ free_area_available_for_new_partitions(context, *b));
}
-static uint64_t charge_size(uint64_t total, uint64_t amount) {
+static uint64_t charge_size(Context *context, uint64_t total, uint64_t amount) {
+ assert(context);
/* Subtract the specified amount from total, rounding up to multiple of 4K if there's room */
assert(amount <= total);
- return LESS_BY(total, round_up_size(amount, 4096));
+ return LESS_BY(total, round_up_size(amount, context->grain_size));
}
static uint64_t charge_weight(uint64_t total, uint64_t amount) {
@@ -545,14 +566,14 @@ static bool context_allocate_partitions(Context *context, uint64_t *ret_largest_
assert(context);
/* Sort free areas by size, putting smallest first */
- typesafe_qsort(context->free_areas, context->n_free_areas, free_area_compare);
+ typesafe_qsort_r(context->free_areas, context->n_free_areas, free_area_compare, context);
/* In any case return size of the largest free area (i.e. not the size of all free areas
* combined!) */
if (ret_largest_free_area)
*ret_largest_free_area =
context->n_free_areas == 0 ? 0 :
- free_area_available_for_new_partitions(context->free_areas[context->n_free_areas-1]);
+ free_area_available_for_new_partitions(context, context->free_areas[context->n_free_areas-1]);
/* A simple first-fit algorithm. We return true if we can fit the partitions in, otherwise false. */
LIST_FOREACH(partitions, p, context->partitions) {
@@ -565,13 +586,13 @@ static bool context_allocate_partitions(Context *context, uint64_t *ret_largest_
continue;
/* How much do we need to fit? */
- required = partition_min_size_with_padding(p);
- assert(required % 4096 == 0);
+ required = partition_min_size_with_padding(context, p);
+ assert(required % context->grain_size == 0);
for (size_t i = 0; i < context->n_free_areas; i++) {
a = context->free_areas[i];
- if (free_area_available_for_new_partitions(a) >= required) {
+ if (free_area_available_for_new_partitions(context, a) >= required) {
fits = true;
break;
}
@@ -683,8 +704,8 @@ static int context_grow_partitions_phase(
if (r < 0)
return r;
- rsz = partition_min_size(p);
- xsz = partition_max_size(p);
+ rsz = partition_min_size(context, p);
+ xsz = partition_max_size(context, p);
if (phase == PHASE_OVERCHARGE && rsz > share) {
/* This partition needs more than its calculated share. Let's assign
@@ -712,13 +733,13 @@ static int context_grow_partitions_phase(
/* Never change of foreign partitions (i.e. those we don't manage) */
p->new_size = p->current_size;
else
- p->new_size = MAX(round_down_size(share, 4096), rsz);
+ p->new_size = MAX(round_down_size(share, context->grain_size), rsz);
charge = true;
}
if (charge) {
- *span = charge_size(*span, p->new_size);
+ *span = charge_size(context, *span, p->new_size);
*weight_sum = charge_weight(*weight_sum, p->weight);
}
@@ -742,7 +763,7 @@ static int context_grow_partitions_phase(
charge = try_again = true;
} else if (phase == PHASE_DISTRIBUTE) {
- p->new_padding = round_down_size(share, 4096);
+ p->new_padding = round_down_size(share, context->grain_size);
if (p->padding_min != UINT64_MAX && p->new_padding < p->padding_min)
p->new_padding = p->padding_min;
@@ -750,7 +771,7 @@ static int context_grow_partitions_phase(
}
if (charge) {
- *span = charge_size(*span, p->new_padding);
+ *span = charge_size(context, *span, p->new_padding);
*weight_sum = charge_weight(*weight_sum, p->padding_weight);
}
@@ -779,7 +800,7 @@ static int context_grow_partitions_on_free_area(Context *context, FreeArea *a) {
assert(a->after->offset != UINT64_MAX);
assert(a->after->current_size != UINT64_MAX);
- span += round_up_size(a->after->offset + a->after->current_size, 4096) - a->after->offset;
+ span += round_up_size(a->after->offset + a->after->current_size, context->grain_size) - a->after->offset;
}
for (GrowPartitionPhase phase = 0; phase < _GROW_PARTITION_PHASE_MAX;) {
@@ -799,13 +820,13 @@ static int context_grow_partitions_on_free_area(Context *context, FreeArea *a) {
assert(a->after->new_size != UINT64_MAX);
/* Calculate new size and align (but ensure this doesn't shrink the size) */
- m = MAX(a->after->new_size, round_down_size(a->after->new_size + span, 4096));
+ m = MAX(a->after->new_size, round_down_size(a->after->new_size + span, context->grain_size));
- xsz = partition_max_size(a->after);
+ xsz = partition_max_size(context, a->after);
if (xsz != UINT64_MAX && m > xsz)
m = xsz;
- span = charge_size(span, m - a->after->new_size);
+ span = charge_size(context, span, m - a->after->new_size);
a->after->new_size = m;
}
@@ -824,13 +845,13 @@ static int context_grow_partitions_on_free_area(Context *context, FreeArea *a) {
continue;
assert(p->new_size != UINT64_MAX);
- m = MAX(p->new_size, round_down_size(p->new_size + span, 4096));
+ m = MAX(p->new_size, round_down_size(p->new_size + span, context->grain_size));
- xsz = partition_max_size(p);
+ xsz = partition_max_size(context, p);
if (xsz != UINT64_MAX && m > xsz)
m = xsz;
- span = charge_size(span, m - p->new_size);
+ span = charge_size(context, span, m - p->new_size);
p->new_size = m;
if (span == 0)
@@ -910,7 +931,7 @@ static void context_place_partitions(Context *context) {
} else
start = context->start;
- start = round_up_size(start, 4096);
+ start = round_up_size(start, context->grain_size);
left = a->size;
LIST_FOREACH(partitions, p, context->partitions) {
@@ -1422,6 +1443,8 @@ static int determine_current_padding(
struct fdisk_context *c,
struct fdisk_table *t,
struct fdisk_partition *p,
+ uint64_t secsz,
+ uint64_t grainsz,
uint64_t *ret) {
size_t n_partitions;
@@ -1435,8 +1458,8 @@ static int determine_current_padding(
return log_error_errno(SYNTHETIC_ERRNO(EIO), "Partition has no end!");
offset = fdisk_partition_get_end(p);
- assert(offset < UINT64_MAX / 512);
- offset *= 512;
+ assert(offset < UINT64_MAX / secsz);
+ offset *= secsz;
n_partitions = fdisk_table_get_nents(t);
for (size_t i = 0; i < n_partitions; i++) {
@@ -1454,8 +1477,8 @@ static int determine_current_padding(
continue;
start = fdisk_partition_get_start(q);
- assert(start < UINT64_MAX / 512);
- start *= 512;
+ assert(start < UINT64_MAX / secsz);
+ start *= secsz;
if (start >= offset && (next == UINT64_MAX || next > start))
next = start;
@@ -1467,16 +1490,16 @@ static int determine_current_padding(
assert(next < UINT64_MAX);
next++; /* The last LBA is one sector before the end */
- assert(next < UINT64_MAX / 512);
- next *= 512;
+ assert(next < UINT64_MAX / secsz);
+ next *= secsz;
if (offset > next)
return log_error_errno(SYNTHETIC_ERRNO(EIO), "Partition end beyond disk end.");
}
assert(next >= offset);
- offset = round_up_size(offset, 4096);
- next = round_down_size(next, 4096);
+ offset = round_up_size(offset, grainsz);
+ next = round_down_size(next, grainsz);
*ret = LESS_BY(next, offset); /* Saturated subtraction, rounding might have fucked things up */
return 0;
@@ -1549,6 +1572,8 @@ static int context_load_partition_table(
bool from_scratch = false;
sd_id128_t disk_uuid;
size_t n_partitions;
+ unsigned long secsz;
+ uint64_t grainsz;
int r;
assert(context);
@@ -1583,8 +1608,12 @@ static int context_load_partition_table(
if (r < 0)
return log_error_errno(errno, "Failed to stat block device '%s': %m", node);
- if (S_ISREG(st.st_mode) && st.st_size == 0)
+ if (S_ISREG(st.st_mode) && st.st_size == 0) {
+ /* User the fallback values if we have no better idea */
+ context->sector_size = 512;
+ context->grain_size = 4096;
return /* from_scratch = */ true;
+ }
r = -EINVAL;
}
@@ -1602,6 +1631,23 @@ static int context_load_partition_table(
if (flock(fdisk_get_devfd(c), arg_dry_run ? LOCK_SH : LOCK_EX) < 0)
return log_error_errno(errno, "Failed to lock block device: %m");
+ /* The offsets/sizes libfdisk returns to us will be in multiple of the sector size of the
+ * device. This is typically 512, and sometimes 4096. Let's query libfdisk once for it, and then use
+ * it for all our needs. Note that the values we use ourselves always are in bytes though, thus mean
+ * the same thing universally. Also note that regardless what kind of sector size is in use we'll
+ * place partitions at multiples of 4K. */
+ secsz = fdisk_get_sector_size(c);
+
+ /* Insist on a power of two, and that it's a multiple of 512, i.e. the traditional sector size. */
+ if (secsz < 512 || secsz != 1UL << log2u64(secsz))
+ return log_error_errno(errno, "Sector size %lu is not a power of two larger than 512? Refusing.", secsz);
+
+ /* Use at least 4K, and ensure it's a multiple of the sector size, regardless if that is smaller or
+ * larger */
+ grainsz = secsz < 4096 ? 4096 : secsz;
+
+ log_debug("Sector size of device is %lu bytes. Using grain size of %" PRIu64 ".", secsz, grainsz);
+
switch (arg_empty) {
case EMPTY_REFUSE:
@@ -1732,12 +1778,12 @@ static int context_load_partition_table(
}
sz = fdisk_partition_get_size(p);
- assert_se(sz <= UINT64_MAX/512);
- sz *= 512;
+ assert_se(sz <= UINT64_MAX/secsz);
+ sz *= secsz;
start = fdisk_partition_get_start(p);
- assert_se(start <= UINT64_MAX/512);
- start *= 512;
+ assert_se(start <= UINT64_MAX/secsz);
+ start *= secsz;
partno = fdisk_partition_get_partno(p);
@@ -1762,7 +1808,7 @@ static int context_load_partition_table(
pp->current_partition = p;
fdisk_ref_partition(p);
- r = determine_current_padding(c, t, p, &pp->current_padding);
+ r = determine_current_padding(c, t, p, secsz, grainsz, &pp->current_padding);
if (r < 0)
return r;
@@ -1795,7 +1841,7 @@ static int context_load_partition_table(
np->current_partition = p;
fdisk_ref_partition(p);
- r = determine_current_padding(c, t, p, &np->current_padding);
+ r = determine_current_padding(c, t, p, secsz, grainsz, &np->current_padding);
if (r < 0)
return r;
@@ -1812,26 +1858,26 @@ static int context_load_partition_table(
add_initial_free_area:
nsectors = fdisk_get_nsectors(c);
- assert(nsectors <= UINT64_MAX/512);
- nsectors *= 512;
+ assert(nsectors <= UINT64_MAX/secsz);
+ nsectors *= secsz;
first_lba = fdisk_get_first_lba(c);
- assert(first_lba <= UINT64_MAX/512);
- first_lba *= 512;
+ assert(first_lba <= UINT64_MAX/secsz);
+ first_lba *= secsz;
last_lba = fdisk_get_last_lba(c);
assert(last_lba < UINT64_MAX);
last_lba++;
- assert(last_lba <= UINT64_MAX/512);
- last_lba *= 512;
+ assert(last_lba <= UINT64_MAX/secsz);
+ last_lba *= secsz;
assert(last_lba >= first_lba);
if (left_boundary == UINT64_MAX) {
/* No partitions at all? Then the whole disk is up for grabs. */
- first_lba = round_up_size(first_lba, 4096);
- last_lba = round_down_size(last_lba, 4096);
+ first_lba = round_up_size(first_lba, grainsz);
+ last_lba = round_down_size(last_lba, grainsz);
if (last_lba > first_lba) {
r = context_add_free_area(context, last_lba - first_lba, NULL);
@@ -1842,9 +1888,9 @@ add_initial_free_area:
/* Add space left of first partition */
assert(left_boundary >= first_lba);
- first_lba = round_up_size(first_lba, 4096);
- left_boundary = round_down_size(left_boundary, 4096);
- last_lba = round_down_size(last_lba, 4096);
+ first_lba = round_up_size(first_lba, grainsz);
+ left_boundary = round_down_size(left_boundary, grainsz);
+ last_lba = round_down_size(last_lba, grainsz);
if (left_boundary > first_lba) {
r = context_add_free_area(context, left_boundary - first_lba, NULL);
@@ -1856,6 +1902,8 @@ add_initial_free_area:
context->start = first_lba;
context->end = last_lba;
context->total = nsectors;
+ context->sector_size = secsz;
+ context->grain_size = grainsz;
context->fdisk_context = TAKE_PTR(c);
return from_scratch;
@@ -2360,7 +2408,7 @@ static int context_discard_range(
if (S_ISBLK(st.st_mode)) {
uint64_t range[2], end;
- range[0] = round_up_size(offset, 512);
+ range[0] = round_up_size(offset, context->sector_size);
if (offset > UINT64_MAX - size)
return -ERANGE;
@@ -2369,7 +2417,7 @@ static int context_discard_range(
if (end <= range[0])
return 0;
- range[1] = round_down_size(end - range[0], 512);
+ range[1] = round_down_size(end - range[0], context->sector_size);
if (range[1] <= 0)
return 0;
@@ -2519,6 +2567,7 @@ static int context_wipe_and_discard(Context *context, bool from_scratch) {
}
static int partition_encrypt(
+ Context *context,
Partition *p,
const char *node,
struct crypt_device **ret_cd,
@@ -2532,6 +2581,7 @@ static int partition_encrypt(
sd_id128_t uuid;
int r;
+ assert(context);
assert(p);
assert(p->encrypt != ENCRYPT_OFF);
@@ -2579,7 +2629,7 @@ static int partition_encrypt(
volume_key_size,
&(struct crypt_params_luks2) {
.label = strempty(p->new_label),
- .sector_size = 512U,
+ .sector_size = context->sector_size,
});
if (r < 0)
return log_error_errno(r, "Failed to LUKS2 format future partition: %m");
@@ -2735,7 +2785,7 @@ static int context_copy_blocks(Context *context) {
if (r < 0)
return log_error_errno(r, "Failed to lock loopback device: %m");
- r = partition_encrypt(p, d->node, &cd, &encrypted, &encrypted_dev_fd);
+ r = partition_encrypt(context, p, d->node, &cd, &encrypted, &encrypted_dev_fd);
if (r < 0)
return log_error_errno(r, "Failed to encrypt device: %m");
@@ -2988,7 +3038,7 @@ static int context_mkfs(Context *context) {
return log_error_errno(r, "Failed to lock loopback device: %m");
if (p->encrypt != ENCRYPT_OFF) {
- r = partition_encrypt(p, d->node, &cd, &encrypted, &encrypted_dev_fd);
+ r = partition_encrypt(context, p, d->node, &cd, &encrypted, &encrypted_dev_fd);
if (r < 0)
return log_error_errno(r, "Failed to encrypt device: %m");
@@ -3307,13 +3357,13 @@ static int context_mangle_partitions(Context *context) {
if (p->new_size != p->current_size) {
assert(p->new_size >= p->current_size);
- assert(p->new_size % 512 == 0);
+ assert(p->new_size % context->sector_size == 0);
r = fdisk_partition_size_explicit(p->current_partition, true);
if (r < 0)
return log_error_errno(r, "Failed to enable explicit sizing: %m");
- r = fdisk_partition_set_size(p->current_partition, p->new_size / 512);
+ r = fdisk_partition_set_size(p->current_partition, p->new_size / context->sector_size);
if (r < 0)
return log_error_errno(r, "Failed to grow partition: %m");
@@ -3353,8 +3403,8 @@ static int context_mangle_partitions(Context *context) {
_cleanup_(fdisk_unref_parttypep) struct fdisk_parttype *t = NULL;
assert(!p->new_partition);
- assert(p->offset % 512 == 0);
- assert(p->new_size % 512 == 0);
+ assert(p->offset % context->sector_size == 0);
+ assert(p->new_size % context->sector_size == 0);
assert(!sd_id128_is_null(p->new_uuid));
assert(p->new_label);
@@ -3378,11 +3428,11 @@ static int context_mangle_partitions(Context *context) {
if (r < 0)
return log_error_errno(r, "Failed to enable explicit sizing: %m");
- r = fdisk_partition_set_start(q, p->offset / 512);
+ r = fdisk_partition_set_start(q, p->offset / context->sector_size);
if (r < 0)
return log_error_errno(r, "Failed to position partition: %m");
- r = fdisk_partition_set_size(q, p->new_size / 512);
+ r = fdisk_partition_set_size(q, p->new_size / context->sector_size);
if (r < 0)
return log_error_errno(r, "Failed to grow partition: %m");
@@ -4746,18 +4796,20 @@ done:
}
static int determine_auto_size(Context *c) {
- uint64_t sum = round_up_size(GPT_METADATA_SIZE, 4096);
+ uint64_t sum;
Partition *p;
assert_se(c);
+ sum = round_up_size(GPT_METADATA_SIZE, 4096);
+
LIST_FOREACH(partitions, p, c->partitions) {
uint64_t m;
if (p->dropped)
continue;
- m = partition_min_size_with_padding(p);
+ m = partition_min_size_with_padding(c, p);
if (m > UINT64_MAX - sum)
return log_error_errno(SYNTHETIC_ERRNO(EOVERFLOW), "Image would grow too large, refusing.");