idrac 0.10.4 → 0.10.6

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 17491587256894c2ff168665dca777903a2d6b7d7582f745c4c3e81f697889cd
-  data.tar.gz: 432ecbfd6df649a69644254dc942e253a8d917f6ce2cd07c30f7f328c3ffee75
+  metadata.gz: fae473e8260d010235ebf86c7ed40b33dae6b7a1c14c2576bce55fecc319c7be
+  data.tar.gz: 8c115aeb8c6399d2d3d167204b160ef691b64ec41570b371912df697b23d9789
 SHA512:
-  metadata.gz: 28cc04b7ec2f246ed29d95ba50481f52dff7893e69fd94dc6f82db81e0d7b988e6d0e21589ae09a75e99c51621f2a6650a1aa0ada2c74c8aa4ff6f2726cadf98
-  data.tar.gz: 6981e38fe25c803372a0c09aa3616db6ce7d0670415fbec9f73e143c25e44fff25e24ea650dfe5ebfcc821b14c6d17dd778664e43d11b5fbfd8c6458cacae374
+  metadata.gz: 278b81e8955021924c44c496b0df20c34f59e3630f9df3decda96ee519cadc0445901cddbf3c78c174313c36eef558780ce444fb09f3d870039adc23e7da5056
+  data.tar.gz: c66f192778e2e32cf724b3b98ef45c2311a79ee8f0144e0f95a3106ff1d2d87f67f13c30eac4a47c2055e302bb5b2dc373cf37cfc760ff2566a8f949531db061

data/lib/idrac/firmware.rb

@@ -11,6 +11,8 @@ require_relative 'firmware_catalog'
 require 'faraday'
 require 'faraday/multipart'
 
+require "timeout"
+
 module IDRAC
   class Firmware
     attr_reader :client
@@ -90,12 +92,29 @@ module IDRAC
               # Id ("Installed-<componentID>-<version>"). Used to match DUPs by
               # componentID rather than by ambiguous display name.
               fw_id = component_data['Id'].to_s
-              component_id = component_data.dig('Oem', 'Dell', 'DellSoftwareInventory', 'ComponentID')
+              oem = component_data.dig('Oem', 'Dell', 'DellSoftwareInventory') || {}
+              component_id = oem['ComponentID']
               component_id ||= fw_id[/\A(?:Installed|Current|Previous|Available)-(\d+)-/, 1]
+              # The same Oem block carries the device's PCI IDs, and it carries
+              # them exactly where the componentID is missing. The iDRAC reports
+              # componentID "0" for firmware it did not install itself, so
+              # factory-flashed hardware is unmatchable by componentID: on an
+              # R6525, 13 of 23 entries report "0" (both LOMs, both ConnectX-6,
+              # the PERC, all 8 NVMe drives) and each publishes all four PCI
+              # IDs, while the 10 with a real componentID publish none. An
+              # entry gains its componentID once a DUP is applied — a LOM went
+              # from "Installed-0-22.91.5" to "Installed-108255-23.61.3" — so
+              # the unmatchable set is first-touch hardware, not an edge case.
+              # No PCIeDevices/PCIeFunctions lookup is needed to identify what
+              # the componentID cannot; see #updates_for_component.
               firmware_inventory << {
                 name: component_data['Name'],
                 id: component_data['Id'],
                 component_id: component_id,
+                vendor_id: oem['VendorID'],
+                device_id: oem['DeviceID'],
+                sub_vendor_id: oem['SubVendorID'],
+                sub_device_id: oem['SubDeviceID'],
                 version: component_data['Version'],
                 updateable: component_data['Updateable'] || false,
                 status: component_data['Status'] ? component_data['Status']['State'] : 'Unknown'
@@ -423,6 +442,51 @@ module IDRAC
 
 
 
+    # An iDRAC holds ONE staged package and refuses another while any update job is outstanding
+    # ("503: A deployment operation is already in progress"). Retrying cannot clear it, because
+    # the staged package IS the operation -- the caller has to wait for the slot. Anyone updating
+    # several components in sequence needs this: a straight loop over one R6525's 15 packages
+    # submitted 6 and lost 9 to 503.
+    #
+    # Every rule here is a Dell fact rather than a caller's policy, which is why it belongs in
+    # the gem. Each was a silent stall on real hardware:
+    #
+    #   * "Available-" is a PREFIX, not a substring. The TPM ships an installed entry whose
+    #     version is the literal string "NotAvailable"
+    #     (Installed-12345-NotAvailable__TPM.Integrated.1-1); a substring test matches it and the
+    #     slot then never reads free on that machine.
+    #   * RebootCompleted is TERMINAL, as is anything at PercentComplete 100. Counting a finished
+    #     reboot as active made a drained queue look busy for 67 minutes.
+    #   * Jobs matter, not just the staging slot: triggering an install consumes the staged
+    #     package into a job, so Available empties while the install is still pending.
+    TERMINAL_JOB_STATES = %w[Completed Failed CompletedWithErrors RebootCompleted RebootFailed].freeze
+
+    def update_slot_free?
+      staged = JSON.parse(client.authenticated_request(:get, "/redfish/v1/UpdateService/FirmwareInventory").body)["Members"]
+                   .map { |m| m["@odata.id"].to_s.split("/").last }
+                   .count { |id| id.start_with?("Available-") }
+      busy = JSON.parse(client.authenticated_request(:get, "/redfish/v1/Managers/iDRAC.Embedded.1/Oem/Dell/Jobs?$expand=*($levels=1)").body)["Members"]
+                 .count { |job| !(TERMINAL_JOB_STATES.include?(job["JobState"]) || job["PercentComplete"].to_i == 100) }
+      staged.zero? && busy.zero?
+    end
+
+    # Block until the iDRAC can accept an update. Probes are bounded because a hung transport
+    # blocks without raising, so a bare rescue never fires and the wait wedges silently.
+    def wait_for_update_slot!(timeout: 2700, interval: 30, probe_timeout: 90)
+      deadline = Time.now + timeout
+      loop do
+        begin
+          return true if Timeout.timeout(probe_timeout) { update_slot_free? }
+        rescue StandardError => e
+          # Report rather than counting a probe failure as "busy": an unreachable BMC is
+          # otherwise indistinguishable from a working one, and the caller waits out the timeout.
+          puts "Update slot probe failed (#{e.class}); retrying".yellow
+        end
+        raise Error, "iDRAC update slot never freed within #{timeout}s" if Time.now > deadline
+        sleep interval
+      end
+    end
+
     private
 
     def upload_firmware(firmware_path)

data/lib/idrac/firmware_catalog.rb

@@ -172,6 +172,16 @@ module IDRAC
         # picks wrong DUPs and iDRAC rejects them (RED097 / "not compatible").
         component_ids = component.xpath(".//SupportedDevices/Device/@componentID").map(&:value).uniq
 
+        # The PCI IDs this DUP declares it can flash. The PCIe/storage half of
+        # the iDRAC inventory reports componentID "0" (measured on an R6525:
+        # every NIC, the PERC and all 8 NVMe drives), so componentID alone
+        # cannot match them and the old name heuristic guessed — it offered a
+        # Broadcom LOM an Intel package. PCI IDs are the identity both sides
+        # publish, so a cross-vendor match becomes impossible.
+        pci_ids = component.xpath(".//SupportedDevices/Device/PCIInfo").map { |p|
+          pci_key(p["vendorID"], p["deviceID"], p["subVendorID"], p["subDeviceID"])
+        }.compact.uniq
+
         # Skip if missing essential information
         next if name.empty? || path.empty? || version.empty?
 
@@ -191,6 +201,7 @@ module IDRAC
             component_type: component_type,
             category: category,
             component_ids: component_ids,
+            pci_ids: pci_ids,
             download_url: "https://downloads.dell.com/#{path}"
           }
         end
@@ -200,24 +211,66 @@ module IDRAC
       updates
     end
 
+    # Normalize one PCI ID quad to a comparable key. Dell publishes the same
+    # four IDs on both sides but spells them differently: the catalog writes
+    # bare uppercase hex (<PCIInfo vendorID="14E4" deviceID="165F"/>) while
+    # Redfish writes "0x790e". Returns nil unless all four are present, so a
+    # caller can tell "no PCI identity" from "PCI identity that matched
+    # nothing".
+    def pci_key(vendor_id, device_id, sub_vendor_id, sub_device_id)
+      parts = [vendor_id, device_id, sub_vendor_id, sub_device_id].map do |id|
+        id.to_s.strip.sub(/\A0x/i, "").upcase
+      end
+      return nil if parts.any?(&:empty?)
+
+      parts.map { |p| p.rjust(4, "0") }.join("/")
+    end
+
     # Select the catalog DUPs that apply to one installed component (`fw` from
-    # the iDRAC inventory). Matches on Dell componentID — the reliable key —
-    # and only falls back to the legacy name heuristic when the inventory
-    # didn't expose a componentID (e.g. it reports "0").
+    # the iDRAC inventory), most specific key first.
+    #
+    # 1. Dell componentID, when the inventory gives a real one. Exact and
+    #    unambiguous; unchanged.
+    # 2. PCI IDs, for the entries that report componentID "0" — on an R6525
+    #    that is every NIC, the PERC and every drive. The full quad first, then
+    #    vendor+device if the catalog lists no package for that exact subsystem
+    #    ID. Both tiers pin the vendor, which is the whole point: name matching
+    #    reduced any NIC to the token "NIC" and any RAID device to "PERC", then
+    #    substring-matched the DUP's package name, so a "Broadcom NetXtreme
+    #    Gigabit Ethernet" LOM was offered "Intel NIC Family Version 25.0.0
+    #    Firmware for Intel I350 and X550 Adapters" — wrong vendor — and a
+    #    "PERC H755N Front" was offered the H355 package. It was also unstable:
+    #    the same tool offered two identical PERCs different packages, because
+    #    the winner depended on catalog order.
+    # 3. The name heuristic, only when the inventory gave neither key. When PCI
+    #    IDs are present they are the answer: if nothing matches them, the
+    #    catalog has no package for this device, and guessing by name is how
+    #    the wrong-vendor DUPs got offered in the first place.
     def updates_for_component(catalog_updates, fw)
       component_id = fw[:component_id]
       if component_id && component_id != "0"
-        catalog_updates.select { |u| Array(u[:component_ids]).include?(component_id) }
-      else
-        name = fw[:name] || ""
-        ids = extract_identifiers(name)
-        catalog_updates.select do |u|
-          un = u[:name] || ""
-          ids.any? { |id| un.downcase.include?(id.downcase) } ||
-            un.downcase.include?(name.downcase) ||
-            name.downcase.include?(un.downcase)
+        return catalog_updates.select { |u| Array(u[:component_ids]).include?(component_id) }
+      end
+
+      key = pci_key(fw[:vendor_id], fw[:device_id], fw[:sub_vendor_id], fw[:sub_device_id])
+      if key
+        exact = catalog_updates.select { |u| Array(u[:pci_ids]).include?(key) }
+        return exact if exact.any?
+
+        chip = key.split("/").first(2).join("/")
+        return catalog_updates.select do |u|
+          Array(u[:pci_ids]).any? { |k| k.start_with?("#{chip}/") }
         end
       end
+
+      name = fw[:name] || ""
+      ids = extract_identifiers(name)
+      catalog_updates.select do |u|
+        un = u[:name] || ""
+        ids.any? { |id| un.downcase.include?(id.downcase) } ||
+          un.downcase.include?(name.downcase) ||
+          name.downcase.include?(un.downcase)
+      end
     end
 
     # Sortable key for a Dell version string (e.g. "25.5.9.0001", "22.00.6").

data/lib/idrac/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module IDRAC
-  VERSION = "0.10.4"
+  VERSION = "0.10.6"
 end

metadata

@@ -1,14 +1,13 @@
 --- !ruby/object:Gem::Specification
 name: idrac
 version: !ruby/object:Gem::Version
-  version: 0.10.4
+  version: 0.10.6
 platform: ruby
 authors:
 - Jonathan Siegel
-autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-08-29 00:00:00.000000000 Z
+date: 1980-01-02 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: httparty
@@ -282,7 +281,6 @@ licenses:
 metadata:
   homepage_uri: https://github.com/buildio/idrac
   source_code_uri: https://github.com/buildio/idrac
-post_install_message:
 rdoc_options: []
 require_paths:
 - lib
@@ -297,8 +295,7 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.5.16
-signing_key:
+rubygems_version: 3.6.9
 specification_version: 4
 summary: API Client for Dell iDRAC
 test_files: []