Fix nRF52 contact persistence and ExtraFS recovery
This commit is contained in:
1 parent
66ad8c52df
commit
cb27fa8539
23 files changed
+1591
-113
No files matched your search
+19
-2
@@ -277,6 +277,14 @@ void Mesh::loop() {
|
||||
#endif
|
||||
}
|
||||
|
||||
bool Mesh::hasPendingOtaApply() const {
|
||||
#if defined(ENABLE_OTA) && !defined(OTA_SEEDER_ONLY)
|
||||
return ota::ota_ctx().apply_pending;
|
||||
#else
|
||||
return false;
|
||||
#endif
|
||||
}
|
||||
|
||||
void __attribute__((noinline)) Mesh::serviceLoopMaintenance() {
|
||||
if (_waiting_direct_retry_count != 0
|
||||
&& millisHasNowPassed(_next_direct_retry_timeout)) {
|
||||
@@ -335,8 +343,17 @@ void __attribute__((noinline)) Mesh::serviceLoopMaintenance() {
|
||||
oc.apply_hard = futureMillis(15000);
|
||||
} else if (millisHasNowPassed(oc.apply_at) &&
|
||||
(_mgr->getOutboundTotal() == 0 || millisHasNowPassed(oc.apply_hard))) {
|
||||
if (oc.bootloader_apply_pending) ota::ota_reboot_to_bootloader_update();
|
||||
else ota::ota_reboot_to_apply();
|
||||
if (!prepareForOtaReboot()) {
|
||||
// A role-specific persistence flush failed or is in bounded
|
||||
// backoff. Keep the verified apply armed without hammering storage
|
||||
// on every pass through the event loop.
|
||||
oc.apply_at = futureMillis(1000);
|
||||
oc.apply_hard = futureMillis(15000);
|
||||
} else if (oc.bootloader_apply_pending) {
|
||||
ota::ota_reboot_to_bootloader_update();
|
||||
} else {
|
||||
ota::ota_reboot_to_apply();
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@@ -190,6 +190,13 @@ class Mesh : public Dispatcher {
|
||||
DispatcherAction forwardMultipartDirect(Packet* pkt);
|
||||
|
||||
protected:
|
||||
// Roles with lazily persisted state can veto an OTA reset until that state
|
||||
// is durable. The OTA apply remains armed and is retried later.
|
||||
virtual bool prepareForOtaReboot() { return true; }
|
||||
// Power-saving roles must stay runnable while the deferred OTA reboot grace
|
||||
// period (or a persistence retry) is pending. On nRF52, board.sleep() has no
|
||||
// timer wake and could otherwise strand an already-approved update.
|
||||
bool hasPendingOtaApply() const;
|
||||
DispatcherAction onRecvPacket(Packet* pkt) override;
|
||||
void onTracePacketQueuedForSend(Packet* packet) override;
|
||||
void onSendComplete(Packet* packet) override;
|
||||
|
||||
@@ -1143,10 +1143,14 @@ void WebConfigServer::tick(uint32_t now) {
|
||||
}
|
||||
|
||||
if (WebConfigBatch::rebootDue(_reboot_at, now)) {
|
||||
// rebootNow() can return when its pre-reboot persistence flush fails.
|
||||
// Consume this request first so a full filesystem cannot turn tick() into
|
||||
// a tight loop of repeated flush attempts.
|
||||
_reboot_at = 0;
|
||||
mesh::usbLoggingPort().printf(
|
||||
"WC: rebooting now (%s)\n",
|
||||
_batch_reboot_armed ? "confirmed" : "fallback");
|
||||
_cb->rebootNow(); // does not return
|
||||
_cb->rebootNow();
|
||||
}
|
||||
|
||||
if ((int32_t)(_diag_until - now) > 0 && (now - _diag_last) >= 1000) {
|
||||
|
||||
@@ -0,0 +1,64 @@
|
||||
#pragma once
|
||||
|
||||
#include <stdint.h>
|
||||
|
||||
namespace mesh {
|
||||
namespace storage {
|
||||
|
||||
static const uint32_t INTERNAL_EXTRAFS_START = 0xD4000UL;
|
||||
static const uint32_t INTERNAL_EXTRAFS_SIZE = 0x19000UL;
|
||||
static const uint32_t INTERNAL_EXTRAFS_BLOCK_SIZE = 128UL;
|
||||
|
||||
inline bool isExpectedInternalExtraFsGeometry(uint32_t start, uint32_t size,
|
||||
uint32_t block_size) {
|
||||
return start == INTERNAL_EXTRAFS_START
|
||||
&& size == INTERNAL_EXTRAFS_SIZE
|
||||
&& block_size == INTERNAL_EXTRAFS_BLOCK_SIZE;
|
||||
}
|
||||
|
||||
inline bool isInternalExtraFsReservedByApplication(uint32_t application_end) {
|
||||
return application_end <= INTERNAL_EXTRAFS_START;
|
||||
}
|
||||
|
||||
enum class InternalSecondaryFsBootResult {
|
||||
Mounted,
|
||||
InitializedBlank,
|
||||
PreservedNonBlank,
|
||||
InitializationFailed,
|
||||
};
|
||||
|
||||
// CustomLFS::begin() automatically formats after a failed mount. Use the base
|
||||
// non-formatting mount through this policy instead: an erased factory region
|
||||
// may be initialized, while nonblank media is preserved for explicit repair.
|
||||
template <typename Mount, typename IsBlank, typename Format>
|
||||
InternalSecondaryFsBootResult prepareInternalSecondaryFilesystem(
|
||||
Mount mount, IsBlank is_blank, Format format) {
|
||||
if (mount()) return InternalSecondaryFsBootResult::Mounted;
|
||||
if (!is_blank()) return InternalSecondaryFsBootResult::PreservedNonBlank;
|
||||
if (!format() || !mount()) {
|
||||
return InternalSecondaryFsBootResult::InitializationFailed;
|
||||
}
|
||||
return InternalSecondaryFsBootResult::InitializedBlank;
|
||||
}
|
||||
|
||||
enum class InternalSecondaryFsRepairResult {
|
||||
Repaired,
|
||||
FormatFailed,
|
||||
MountFailed,
|
||||
ValidationFailed,
|
||||
};
|
||||
|
||||
// Repair is deliberately separate from the normal boot path. A filesystem
|
||||
// which mounts but fails a full metadata traversal may still be forensically
|
||||
// recoverable, so erasing it requires an explicit local operator action.
|
||||
template <typename Format, typename Mount, typename Validate>
|
||||
InternalSecondaryFsRepairResult repairInternalSecondaryFilesystem(
|
||||
Format format, Mount mount, Validate validate) {
|
||||
if (!format()) return InternalSecondaryFsRepairResult::FormatFailed;
|
||||
if (!mount()) return InternalSecondaryFsRepairResult::MountFailed;
|
||||
if (!validate()) return InternalSecondaryFsRepairResult::ValidationFailed;
|
||||
return InternalSecondaryFsRepairResult::Repaired;
|
||||
}
|
||||
|
||||
} // namespace storage
|
||||
} // namespace mesh
|
||||
@@ -87,12 +87,10 @@ bool ota_apply_verify_slot(ApplyState& st) {
|
||||
return st.slot_ok;
|
||||
}
|
||||
|
||||
bool ota_apply_commit() {
|
||||
bool ota_apply_arm() {
|
||||
const esp_partition_t* p = esp_ota_get_next_update_partition(nullptr);
|
||||
if (!p) return false;
|
||||
if (esp_ota_set_boot_partition(p) != ESP_OK) return false;
|
||||
esp_restart(); // does not return
|
||||
return true;
|
||||
return esp_ota_set_boot_partition(p) == ESP_OK;
|
||||
}
|
||||
|
||||
// --- detools callback context -----------------------------------------------------------------
|
||||
@@ -457,7 +455,7 @@ static bool parse_inplace_patch_dims(const uint8_t* payload, uint32_t payload_le
|
||||
bool ota_apply_slot_info(uint32_t*, uint32_t*) { return false; }
|
||||
bool ota_apply_set_manifest(const uint8_t*, uint32_t, const SignerAllowlist&, ApplyState& st) { st = ApplyState(); return false; }
|
||||
bool ota_apply_verify_slot(ApplyState&) { return false; }
|
||||
bool ota_apply_commit() { return false; }
|
||||
bool ota_apply_arm() { return false; }
|
||||
bool ota_apply_detools_mota(const uint8_t*, uint32_t, const SignerAllowlist&, ApplyState& st, char* msg) { st = ApplyState(); strcpy(msg, "use ota_apply_mota_nrf52"); return false; }
|
||||
|
||||
void ota_reboot_to_apply() { // public: set the apply magic + reset (does not return)
|
||||
@@ -1092,7 +1090,7 @@ bool ota_prepare_bootloader_update_nrf52(OtaStoreFlashNrf52& store,
|
||||
bool ota_apply_slot_info(uint32_t*, uint32_t*) { return false; }
|
||||
bool ota_apply_set_manifest(const uint8_t*, uint32_t, const SignerAllowlist&, ApplyState& st) { st = ApplyState(); return false; }
|
||||
bool ota_apply_verify_slot(ApplyState&) { return false; }
|
||||
bool ota_apply_commit() { return false; }
|
||||
bool ota_apply_arm() { return false; }
|
||||
bool ota_apply_detools_mota(const uint8_t*, uint32_t, const SignerAllowlist&, ApplyState& st, char* msg) { st = ApplyState(); strcpy(msg, "unsupported"); return false; }
|
||||
bool ota_apply_mota_nrf52(const uint8_t*, uint32_t, const SignerAllowlist&, ApplyState& st, char* msg) { st = ApplyState(); strcpy(msg, "unsupported"); return false; }
|
||||
bool ota_rescue_mota_nrf52(const uint8_t*, uint32_t, const SignerAllowlist&, const uint8_t*,
|
||||
|
||||
@@ -54,7 +54,10 @@ bool ota_apply_slot_info(uint32_t* addr, uint32_t* size); // the
|
||||
bool ota_apply_set_manifest(const uint8_t* mf, uint32_t len,
|
||||
const SignerAllowlist& allow, ApplyState& st); // parse + verify signature
|
||||
bool ota_apply_verify_slot(ApplyState& st); // hash the slot vs image_hash
|
||||
bool ota_apply_commit(); // set-boot + reboot (no return)
|
||||
// Select the verified A/B slot for the next boot without rebooting inline. The
|
||||
// caller must arm OtaContext::apply_pending so Mesh can drain the command reply
|
||||
// and run its role-specific persistence guard before ota_reboot_to_apply().
|
||||
bool ota_apply_arm();
|
||||
|
||||
// Apply an ESP32 A/B `.mota` (CODEC_DETOOLS_SEQUENTIAL delta or CODEC_FULL image) and arm the inactive
|
||||
// slot; the caller reboots after the confirmation reply. `msg` (>=80 bytes) receives a human-readable
|
||||
|
||||
@@ -1160,8 +1160,15 @@ static bool handle_dev(const char* d, char* reply, OtaContext& c) {
|
||||
sprintf(reply, "slot image_hash %s (size=%u)", ok ? "MATCH" : "MISMATCH", (unsigned)c.apply_st.image_size);
|
||||
} else if (strncmp(sub, "commit", 6) == 0) {
|
||||
if (!c.apply_st.slot_ok) { strcpy(reply, "ERR run 'ota dev apply verify' first (slot must match)"); return true; }
|
||||
ota_apply_commit(); // set boot partition + reboot; no return
|
||||
strcpy(reply, "ERR commit failed (no A/B slot?)");
|
||||
if (c.apply_pending) { strcpy(reply, "ERR update is already armed; reboot is pending"); return true; }
|
||||
if (!ota_apply_arm()) { strcpy(reply, "ERR commit failed (no A/B slot?)"); return true; }
|
||||
// Use the same deferred path as ordinary installs: it lets this reply
|
||||
// drain, then Mesh invokes prepareForOtaReboot() before the reset.
|
||||
c.bootloader_apply_pending = false;
|
||||
c.apply_at = 0;
|
||||
c.apply_hard = 0;
|
||||
c.apply_pending = true;
|
||||
strcpy(reply, "OK slot armed; guarded reboot pending");
|
||||
} else {
|
||||
strcpy(reply, "ERR ota dev apply (slot|manifest|verify|commit)");
|
||||
}
|
||||
|
||||
Reference in new issue
Block a user