aboutsummaryrefslogtreecommitdiff
path: root/gdb/target-memory.c
diff options
context:
space:
mode:
authorDaniel Jacobowitz <drow@false.org>2006-09-21 14:00:53 +0000
committerDaniel Jacobowitz <drow@false.org>2006-09-21 14:00:53 +0000
commita76d924dffcb040b44a2bb5be026f0c974590c30 (patch)
treea347bcb3680f48f5316b790bae5ff8a4105ccf96 /gdb/target-memory.c
parentfd79eceebf094938376c671ea3538a31d4f63eac (diff)
* Makefile.in (SFILES): Add target-memory.c.
(COMMON_OBS): Add target-memory.o. * memattr.c (lookup_mem_region): Adjust handling for the top of memory. Improve comments. * remote.c (packet_check_result): New function, split out from packet_ok. Recognize "E." as an error prefix. (packet_ok): Use it. (remote_write_bytes_aux): New function, renamed from remote_write_bytes. Take packet header, packet format, and length flag as arguments. (remote_write_bytes): Rewrite to use remote_write_bytes_aux. (remote_send_printf, restore_remote_timeout) (remote_flash_timeout, remote_flash_erase, remote_flash_write) (remote_flash_done): New. (remote_xfer_partial): Handle flash writes. (init_remote_ops, init_remote_async_ops): Set to_flash_erase and to_flash_done. * symfile.c (struct load_section_data): Include a pointer to the cumulative stats and a request queue. Move most members to other types. (struct load_progress_data, struct load_progress_section_data): New types. (load_progress): Handle a NULL baton and zero bytes. Update for type changes. (load_section_callback): Create memory write requests instead of writing to memory. Don't print the progress message here. (clear_memory_write_data): New function. (generic_load): Use target_write_memory_blocks. * target-memory.c: New file. * target.c (update_current_target): Mention new uninherited methods. (memory_xfer_partial): Issue an error for flash writes. (target_flash_erase, target_flash_done): New functions. (target_write_with_progress): Call the progress callback at the start also. * target.h (enum target_object): Add TARGET_OBJECT_FLASH. (target_write_with_progress): Update comment. (struct target_ops): Add to_flash_erase and to_flash_done. (target_flash_erase, target_flash_done, struct memory_write_request) (memory_write_request_s, enum flash_preserve_mode) (target_write_memory_blocks): New, including a vector type for memory_write_request_s.
Diffstat (limited to 'gdb/target-memory.c')
-rw-r--r--gdb/target-memory.c437
1 files changed, 437 insertions, 0 deletions
diff --git a/gdb/target-memory.c b/gdb/target-memory.c
new file mode 100644
index 0000000000..64c16e1ac0
--- /dev/null
+++ b/gdb/target-memory.c
@@ -0,0 +1,437 @@
+/* Parts of target interface that deal with accessing memory and memory-like
+ objects.
+
+ Copyright (C) 2006
+ Free Software Foundation, Inc.
+
+ This file is part of GDB.
+
+ This program is free software; you can redistribute it and/or modify
+ it under the terms of the GNU General Public License as published by
+ the Free Software Foundation; either version 2 of the License, or
+ (at your option) any later version.
+
+ This program is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ GNU General Public License for more details.
+
+ You should have received a copy of the GNU General Public License
+ along with this program; if not, write to the Free Software
+ Foundation, Inc., 51 Franklin Street, Fifth Floor,
+ Boston, MA 02110-1301, USA. */
+
+#include "defs.h"
+#include "vec.h"
+#include "target.h"
+#include "memory-map.h"
+
+#include "gdb_assert.h"
+
+#include <stdio.h>
+#include <sys/time.h>
+
+static int
+compare_block_starting_address (const void *a, const void *b)
+{
+ const struct memory_write_request *a_req = a;
+ const struct memory_write_request *b_req = b;
+
+ if (a_req->begin < b_req->begin)
+ return -1;
+ else if (a_req->begin == b_req->begin)
+ return 0;
+ else
+ return 1;
+}
+
+/* Adds to RESULT all memory write requests from BLOCK that are
+ in [BEGIN, END) range.
+
+ If any memory request is only partially in the specified range,
+ that part of the memory request will be added. */
+
+static void
+claim_memory (VEC(memory_write_request_s) *blocks,
+ VEC(memory_write_request_s) **result,
+ ULONGEST begin,
+ ULONGEST end)
+{
+ int i;
+ ULONGEST claimed_begin;
+ ULONGEST claimed_end;
+ struct memory_write_request *r;
+
+ for (i = 0; VEC_iterate (memory_write_request_s, blocks, i, r); ++i)
+ {
+ /* If the request doesn't overlap [BEGIN, END), skip it. We
+ must handle END == 0 meaning the top of memory; we don't yet
+ check for R->end == 0, which would also mean the top of
+ memory, but there's an assertion in
+ target_write_memory_blocks which checks for that. */
+
+ if (begin >= r->end)
+ continue;
+ if (end != 0 && end <= r->begin)
+ continue;
+
+ claimed_begin = max (begin, r->begin);
+ if (end == 0)
+ claimed_end = r->end;
+ else
+ claimed_end = min (end, r->end);
+
+ if (claimed_begin == r->begin && claimed_end == r->end)
+ VEC_safe_push (memory_write_request_s, *result, r);
+ else
+ {
+ struct memory_write_request *n =
+ VEC_safe_push (memory_write_request_s, *result, NULL);
+ memset (n, 0, sizeof (struct memory_write_request));
+ n->begin = claimed_begin;
+ n->end = claimed_end;
+ n->data = r->data + (claimed_begin - r->begin);
+ }
+ }
+}
+
+/* Given a vector of struct memory_write_request objects in BLOCKS,
+ add memory requests for flash memory into FLASH_BLOCKS, and for
+ regular memory to REGULAR_BLOCKS. */
+
+static void
+split_regular_and_flash_blocks (VEC(memory_write_request_s) *blocks,
+ VEC(memory_write_request_s) **regular_blocks,
+ VEC(memory_write_request_s) **flash_blocks)
+{
+ struct mem_region *region;
+ CORE_ADDR cur_address;
+
+ /* This implementation runs in O(length(regions)*length(blocks)) time.
+ However, in most cases the number of blocks will be small, so this does
+ not matter.
+
+ Note also that it's extremely unlikely that a memory write request
+ will span more than one memory region, however for safety we handle
+ such situations. */
+
+ cur_address = 0;
+ while (1)
+ {
+ VEC(memory_write_request_s) **r;
+ region = lookup_mem_region (cur_address);
+
+ r = region->attrib.mode == MEM_FLASH ? flash_blocks : regular_blocks;
+ cur_address = region->hi;
+ claim_memory (blocks, r, region->lo, region->hi);
+
+ if (cur_address == 0)
+ break;
+ }
+}
+
+/* Given an ADDRESS, if BEGIN is non-NULL this function sets *BEGIN
+ to the start of the flash block containing the address. Similarly,
+ if END is non-NULL *END will be set to the address one past the end
+ of the block containing the address. */
+
+static void
+block_boundaries (CORE_ADDR address, CORE_ADDR *begin, CORE_ADDR *end)
+{
+ struct mem_region *region;
+ unsigned blocksize;
+
+ region = lookup_mem_region (address);
+ gdb_assert (region->attrib.mode == MEM_FLASH);
+ blocksize = region->attrib.blocksize;
+ if (begin)
+ *begin = address / blocksize * blocksize;
+ if (end)
+ *end = (address + blocksize - 1) / blocksize * blocksize;
+}
+
+/* Given the list of memory requests to be WRITTEN, this function
+ returns write requests covering each group of flash blocks which must
+ be erased. */
+
+static VEC(memory_write_request_s) *
+blocks_to_erase (VEC(memory_write_request_s) *written)
+{
+ unsigned i;
+ struct memory_write_request *ptr;
+
+ VEC(memory_write_request_s) *result = NULL;
+
+ for (i = 0; VEC_iterate (memory_write_request_s, written, i, ptr); ++i)
+ {
+ CORE_ADDR begin, end;
+
+ block_boundaries (ptr->begin, &begin, 0);
+ block_boundaries (ptr->end, 0, &end);
+
+ if (!VEC_empty (memory_write_request_s, result)
+ && VEC_last (memory_write_request_s, result)->end >= begin)
+ {
+ VEC_last (memory_write_request_s, result)->end = end;
+ }
+ else
+ {
+ struct memory_write_request *n =
+ VEC_safe_push (memory_write_request_s, result, NULL);
+ memset (n, 0, sizeof (struct memory_write_request));
+ n->begin = begin;
+ n->end = end;
+ }
+ }
+
+ return result;
+}
+
+/* Given ERASED_BLOCKS, a list of blocks that will be erased with
+ flash erase commands, and WRITTEN_BLOCKS, the list of memory
+ addresses that will be written, compute the set of memory addresses
+ that will be erased but not rewritten (e.g. padding within a block
+ which is only partially filled by "load"). */
+
+static VEC(memory_write_request_s) *
+compute_garbled_blocks (VEC(memory_write_request_s) *erased_blocks,
+ VEC(memory_write_request_s) *written_blocks)
+{
+ VEC(memory_write_request_s) *result = NULL;
+
+ unsigned i, j;
+ unsigned je = VEC_length (memory_write_request_s, written_blocks);
+ struct memory_write_request *erased_p;
+
+ /* Look at each erased memory_write_request in turn, and
+ see what part of it is subsequently written to.
+
+ This implementation is O(length(erased) * length(written)). If
+ the lists are sorted at this point it could be rewritten more
+ efficiently, but the complexity is not generally worthwhile. */
+
+ for (i = 0;
+ VEC_iterate (memory_write_request_s, erased_blocks, i, erased_p);
+ ++i)
+ {
+ /* Make a deep copy -- it will be modified inside the loop, but
+ we don't want to modify original vector. */
+ struct memory_write_request erased = *erased_p;
+
+ for (j = 0; j != je;)
+ {
+ struct memory_write_request *written
+ = VEC_index (memory_write_request_s,
+ written_blocks, j);
+
+ /* Now try various cases. */
+
+ /* If WRITTEN is fully to the left of ERASED, check the next
+ written memory_write_request. */
+ if (written->end <= erased.begin)
+ {
+ ++j;
+ continue;
+ }
+
+ /* If WRITTEN is fully to the right of ERASED, then ERASED
+ is not written at all. WRITTEN might affect other
+ blocks. */
+ if (written->begin >= erased.end)
+ {
+ VEC_safe_push (memory_write_request_s, result, &erased);
+ goto next_erased;
+ }
+
+ /* If all of ERASED is completely written, we can move on to
+ the next erased region. */
+ if (written->begin <= erased.begin
+ && written->end >= erased.end)
+ {
+ goto next_erased;
+ }
+
+ /* If there is an unwritten part at the beginning of ERASED,
+ then we should record that part and try this inner loop
+ again for the remainder. */
+ if (written->begin > erased.begin)
+ {
+ struct memory_write_request *n =
+ VEC_safe_push (memory_write_request_s, result, NULL);
+ memset (n, 0, sizeof (struct memory_write_request));
+ n->begin = erased.begin;
+ n->end = written->begin;
+ erased.begin = written->begin;
+ continue;
+ }
+
+ /* If there is an unwritten part at the end of ERASED, we
+ forget about the part that was written to and wait to see
+ if the next write request writes more of ERASED. We can't
+ push it yet. */
+ if (written->end < erased.end)
+ {
+ erased.begin = written->end;
+ ++j;
+ continue;
+ }
+ }
+
+ /* If we ran out of write requests without doing anything about
+ ERASED, then that means it's really erased. */
+ VEC_safe_push (memory_write_request_s, result, &erased);
+
+ next_erased:
+ ;
+ }
+
+ return result;
+}
+
+static void
+cleanup_request_data (void *p)
+{
+ VEC(memory_write_request_s) **v = p;
+ struct memory_write_request *r;
+ int i;
+
+ for (i = 0; VEC_iterate (memory_write_request_s, *v, i, r); ++i)
+ xfree (r->data);
+}
+
+static void
+cleanup_write_requests_vector (void *p)
+{
+ VEC(memory_write_request_s) **v = p;
+ VEC_free (memory_write_request_s, *v);
+}
+
+int
+target_write_memory_blocks (VEC(memory_write_request_s) *requests,
+ enum flash_preserve_mode preserve_flash_p,
+ void (*progress_cb) (ULONGEST, void *))
+{
+ struct cleanup *back_to = make_cleanup (null_cleanup, NULL);
+ VEC(memory_write_request_s) *blocks = VEC_copy (memory_write_request_s,
+ requests);
+ unsigned i;
+ int err = 0;
+ struct memory_write_request *r;
+ VEC(memory_write_request_s) *regular = NULL;
+ VEC(memory_write_request_s) *flash = NULL;
+ VEC(memory_write_request_s) *erased, *garbled;
+
+ /* END == 0 would represent wraparound: a write to the very last
+ byte of the address space. This file was not written with that
+ possibility in mind. This is fixable, but a lot of work for a
+ rare problem; so for now, fail noisily here instead of obscurely
+ later. */
+ for (i = 0; VEC_iterate (memory_write_request_s, requests, i, r); ++i)
+ gdb_assert (r->end != 0);
+
+ make_cleanup (cleanup_write_requests_vector, &blocks);
+
+ /* Sort the blocks by their start address. */
+ qsort (VEC_address (memory_write_request_s, blocks),
+ VEC_length (memory_write_request_s, blocks),
+ sizeof (struct memory_write_request), compare_block_starting_address);
+
+ /* Split blocks into list of regular memory blocks,
+ and list of flash memory blocks. */
+ make_cleanup (cleanup_write_requests_vector, &regular);
+ make_cleanup (cleanup_write_requests_vector, &flash);
+ split_regular_and_flash_blocks (blocks, &regular, &flash);
+
+ /* If a variable is added to forbid flash write, even during "load",
+ it should be checked here. Similarly, if this function is used
+ for other situations besides "load" in which writing to flash
+ is undesirable, that should be checked here. */
+
+ /* Find flash blocks to erase. */
+ erased = blocks_to_erase (flash);
+ make_cleanup (cleanup_write_requests_vector, &erased);
+
+ /* Find what flash regions will be erased, and not overwritten; then
+ either preserve or discard the old contents. */
+ garbled = compute_garbled_blocks (erased, flash);
+ make_cleanup (cleanup_request_data, &garbled);
+ make_cleanup (cleanup_write_requests_vector, &garbled);
+
+ if (!VEC_empty (memory_write_request_s, garbled))
+ {
+ if (preserve_flash_p == flash_preserve)
+ {
+ struct memory_write_request *r;
+
+ /* Read in regions that must be preserved and add them to
+ the list of blocks we read. */
+ for (i = 0; VEC_iterate (memory_write_request_s, garbled, i, r); ++i)
+ {
+ gdb_assert (r->data == NULL);
+ r->data = xmalloc (r->end - r->begin);
+ err = target_read_memory (r->begin, r->data, r->end - r->begin);
+ if (err != 0)
+ goto out;
+
+ VEC_safe_push (memory_write_request_s, flash, r);
+ }
+
+ qsort (VEC_address (memory_write_request_s, flash),
+ VEC_length (memory_write_request_s, flash),
+ sizeof (struct memory_write_request), compare_block_starting_address);
+ }
+ }
+
+ /* We could coalesce adjacent memory blocks here, to reduce the
+ number of write requests for small sections. However, we would
+ have to reallocate and copy the data pointers, which could be
+ large; large sections are more common in loadable objects than
+ large numbers of small sections (although the reverse can be true
+ in object files). So, we issue at least one write request per
+ passed struct memory_write_request. The remote stub will still
+ have the opportunity to batch flash requests. */
+
+ /* Write regular blocks. */
+ for (i = 0; VEC_iterate (memory_write_request_s, regular, i, r); ++i)
+ {
+ LONGEST len;
+
+ len = target_write_with_progress (&current_target,
+ TARGET_OBJECT_MEMORY, NULL,
+ r->data, r->begin, r->end - r->begin,
+ progress_cb, r->baton);
+ if (len < (LONGEST) (r->end - r->begin))
+ {
+ /* Call error? */
+ err = -1;
+ goto out;
+ }
+ }
+
+ if (!VEC_empty (memory_write_request_s, erased))
+ {
+ /* Erase all pages. */
+ for (i = 0; VEC_iterate (memory_write_request_s, erased, i, r); ++i)
+ target_flash_erase (r->begin, r->end - r->begin);
+
+ /* Write flash data. */
+ for (i = 0; VEC_iterate (memory_write_request_s, flash, i, r); ++i)
+ {
+ LONGEST len;
+
+ len = target_write_with_progress (&current_target,
+ TARGET_OBJECT_FLASH, NULL,
+ r->data, r->begin, r->end - r->begin,
+ progress_cb, r->baton);
+ if (len < (LONGEST) (r->end - r->begin))
+ error (_("Error writing data to flash"));
+ }
+
+ target_flash_done ();
+ }
+
+ out:
+ do_cleanups (back_to);
+
+ return err;
+}