Proper support for initializing the isolator
Signed-off-by: Zane Shelley <zshelle@us.ibm.com>
Change-Id: Id9a65dc6318420144a7ba9fb78d8a317f1181917
diff --git a/analyzer/analyzer_main.cpp b/analyzer/analyzer_main.cpp
index 774fb0a..cf3d195 100644
--- a/analyzer/analyzer_main.cpp
+++ b/analyzer/analyzer_main.cpp
@@ -4,6 +4,7 @@
#include <hei_main.hpp>
#include <phosphor-logging/log.hpp>
+#include <util/pdbg.hpp>
#include <util/trace.hpp>
#include <algorithm>
@@ -15,50 +16,11 @@
namespace analyzer
{
-/**
- * @brief send chip data file to isolator
- *
- * Read a chip data file into memory and then send it to the isolator via
- * the initialize interface.
- *
- * @param i_filePath The file path and name to read into memory
- *
- * @return Returns true if the isolator was successfully initialized with
- * a single chip data file. Returns false otherwise.
- *
- */
-void initWithFile(const char* i_filePath)
-{
- // open the file and seek to the end to get length
- std::ifstream fileStream(i_filePath, std::ios::binary | std::ios::ate);
+//------------------------------------------------------------------------------
- if (!fileStream.good())
- {
- trace::err("Unable to open file: %s", i_filePath);
- assert(0);
- }
- else
- {
- // get file size based on seek position
- fileStream.seekg(0, std::ios::end);
- std::ifstream::pos_type fileSize = fileStream.tellg();
+// Forward references for externally defined functions.
- // create a buffer large enough to hold the entire file
- std::vector<char> fileBuffer(fileSize);
-
- // seek to the beginning of the file
- fileStream.seekg(0, std::ios::beg);
-
- // read the entire file into the buffer
- fileStream.read(fileBuffer.data(), fileSize);
-
- // done with the file
- fileStream.close();
-
- // initialize the isolator with the chip data
- libhei::initialize(fileBuffer.data(), fileSize);
- }
-}
+void initializeIsolator(const std::vector<libhei::Chip>& i_chips);
//------------------------------------------------------------------------------
@@ -78,11 +40,6 @@
//------------------------------------------------------------------------------
-const char* __path(const libhei::Chip& i_chip)
-{
- return pdbg_target_path((pdbg_target*)i_chip.getChip());
-}
-
const char* __attn(libhei::AttentionType_t i_attnType)
{
const char* str = "";
@@ -130,10 +87,8 @@
//------------------------------------------------------------------------------
-// Returns the chip model/level of the given target. Also, adds the chip
-// model/level to the list of type types needed to initialize the isolator.
-libhei::ChipType_t __getChipType(pdbg_target* i_trgt,
- std::vector<libhei::ChipType_t>& o_types)
+// Returns the chip model/level of the given target.
+libhei::ChipType_t __getChipType(pdbg_target* i_trgt)
{
libhei::ChipType_t type;
@@ -158,23 +113,13 @@
}
// END WORKAROUND
- // Make sure the model/level list contains unique values only.
- // This is O(n*n), but the list size will likely be very low, probably
- // maxing around a half dozen. So, opting for simplicity.
- if (o_types.end() == std::find(o_types.begin(), o_types.end(), type))
- {
- o_types.push_back(type);
- }
-
return type;
}
//------------------------------------------------------------------------------
-// Gathers list of active chips to analyze. Also, returns the list of chip types
-// needed to initialize the isolator.
-void __getActiveChips(std::vector<libhei::Chip>& o_chips,
- std::vector<libhei::ChipType_t>& o_types)
+// Gathers list of active chips to analyze.
+void __getActiveChips(std::vector<libhei::Chip>& o_chips)
{
// Iterate each processor.
pdbg_target* procTrgt;
@@ -185,7 +130,7 @@
continue;
// Add the processor to the list.
- o_chips.emplace_back(procTrgt, __getChipType(procTrgt, o_types));
+ o_chips.emplace_back(procTrgt, __getChipType(procTrgt));
// Iterate the connected OCMBs, if they exist.
pdbg_target* ocmbTrgt;
@@ -196,46 +141,9 @@
continue;
// Add the OCMB to the list.
- o_chips.emplace_back(ocmbTrgt, __getChipType(ocmbTrgt, o_types));
+ o_chips.emplace_back(ocmbTrgt, __getChipType(ocmbTrgt));
}
}
-
- // For debug, trace out all of the chips found.
- for (const auto& chip : o_chips)
- {
- trace::inf("chip:%s type:0x%0" PRIx32, __path(chip), chip.getType());
- }
-}
-
-//------------------------------------------------------------------------------
-
-// Initializes the isolator for each specified chip type.
-void __initializeIsolator(const std::vector<libhei::ChipType_t>& i_types)
-{
- // START WORKAROUND
- // TODO: The chip data will eventually come from the CHIPDATA section of the
- // PNOR. Until that support is available, we'll use temporary chip
- // data files.
- for (const auto& type : i_types)
- {
- switch (type)
- {
- case 0x120DA049: // PROC
- initWithFile(
- "/usr/share/openpower-hw-diags/chip_data_proc.cdb");
- break;
-
- case 0x160D2000: // OCMB_CHIP
- initWithFile(
- "/usr/share/openpower-hw-diags/chip_data_ocmb.cdb");
- break;
-
- default:
- trace::err("Unsupported ChipType_t value: 0x%0" PRIx32, type);
- assert(0);
- }
- }
- // END WORKAROUND
}
//------------------------------------------------------------------------------
@@ -248,8 +156,9 @@
// For debug, trace out the original list of signatures before filtering.
for (const auto& sig : io_list)
{
- trace::inf("Signature: %s 0x%0" PRIx32 " %s", __path(sig.getChip()),
- __sig(sig), __attn(sig.getAttnType()));
+ trace::inf("Signature: %s 0x%0" PRIx32 " %s",
+ util::pdbg::getPath(sig.getChip()), __sig(sig),
+ __attn(sig.getAttnType()));
}
// Special and host attentions are not supported by this user application.
@@ -309,7 +218,8 @@
word8 = __sig(root);
trace::inf("Root cause attention: %s 0x%0" PRIx32 " %s",
- __path(root.getChip()), word8, __attn(root.getAttnType()));
+ util::pdbg::getPath(root.getChip()), word8,
+ __attn(root.getAttnType()));
}
// Get the log data.
@@ -345,19 +255,18 @@
trace::inf(">>> enter analyzeHardware()");
- // Get the active chips to be analyzed and their types.
- std::vector<libhei::Chip> chipList;
- std::vector<libhei::ChipType_t> chipTypes;
- __getActiveChips(chipList, chipTypes);
+ // Get the active chips to be analyzed.
+ std::vector<libhei::Chip> chips;
+ __getActiveChips(chips);
- // Initialize the isolator for all chip types.
- trace::inf("Initializing isolator: # of types=%u", chipTypes.size());
- __initializeIsolator(chipTypes);
+ // Initialize the isolator for all chips.
+ trace::inf("Initializing the isolator...");
+ initializeIsolator(chips);
// Isolate attentions.
- trace::inf("Isolating errors: # of chips=%u", chipList.size());
+ trace::inf("Isolating errors: # of chips=%u", chips.size());
libhei::IsolationData isoData{};
- libhei::isolate(chipList, isoData);
+ libhei::isolate(chips, isoData);
// Filter signatures to determine root cause. We'll need to make a copy of
// the list so that the original list is maintained for the log.
@@ -368,7 +277,7 @@
attnFound = __logError(sigList, isoData);
// All done, clean up the isolator.
- trace::inf("Uninitializing isolator");
+ trace::inf("Uninitializing isolator...");
libhei::uninitialize();
trace::inf("<<< exit analyzeHardware()");
diff --git a/analyzer/initialize_isolator.cpp b/analyzer/initialize_isolator.cpp
new file mode 100644
index 0000000..941b500
--- /dev/null
+++ b/analyzer/initialize_isolator.cpp
@@ -0,0 +1,157 @@
+
+#include <assert.h>
+
+#include <hei_main.hpp>
+#include <util/pdbg.hpp>
+#include <util/trace.hpp>
+
+#include <filesystem>
+#include <fstream>
+#include <map>
+#include <vector>
+
+namespace fs = std::filesystem;
+
+namespace analyzer
+{
+
+//------------------------------------------------------------------------------
+
+void __getChipDataFiles(std::map<libhei::ChipType_t, fs::path>& o_files)
+{
+ o_files.clear();
+
+ auto directory = "/usr/share/openpower-libhei/";
+
+ for (const auto& entry : fs::directory_iterator(directory))
+ {
+ auto path = entry.path();
+
+ std::ifstream file{path, std::ios::binary};
+ if (!file.good())
+ {
+ trace::err("Unable to open file: %s", path.string().c_str());
+ continue;
+ }
+
+ // The first 8-bytes is the file keyword and the next 4-bytes is the
+ // chip type.
+ libhei::FileKeyword_t keyword;
+ libhei::ChipType_t chipType;
+
+ const size_t sz_keyword = sizeof(keyword);
+ const size_t sz_chipType = sizeof(chipType);
+ const size_t sz_buffer = sz_keyword + sz_chipType;
+
+ // Read the keyword and chip type from the file.
+ char buffer[sz_buffer];
+ file.read(buffer, sz_buffer);
+ if (!file.good())
+ {
+ trace::err("Unable to read file: %s", path.string().c_str());
+ continue;
+ }
+
+ // Get the keyword.
+ memcpy(&keyword, &buffer[0], sz_keyword);
+ keyword = be64toh(keyword);
+
+ // Ensure the keyword value is correct.
+ if (libhei::KW_CHIPDATA != keyword)
+ {
+ trace::err("Invalid chip data file: %s", path.string().c_str());
+ continue;
+ }
+
+ // Get the chip type.
+ memcpy(&chipType, &buffer[sz_keyword], sz_chipType);
+ chipType = be32toh(chipType);
+
+ // Trace each legitimate chip data file for debug.
+ trace::inf("File found: type=0x%0" PRIx32 " path=%s", chipType,
+ path.string().c_str());
+
+ // So far, so good. Add the entry.
+ auto ret = o_files.emplace(chipType, path);
+ assert(ret.second); // Should not have duplicate entries
+ }
+}
+
+//------------------------------------------------------------------------------
+
+void __initialize(const fs::path& i_path)
+{
+ // Get file size.
+ const auto sz_buffer = fs::file_size(i_path);
+
+ // Create a buffer large enough to hold the entire file.
+ std::vector<char> buffer(sz_buffer);
+
+ // Open the chip data file.
+ std::ifstream file{i_path, std::ios::binary};
+ assert(file.good()); // We've opened it once before, so it should open now.
+
+ // Read the entire file into the buffer.
+ file.read(buffer.data(), sz_buffer);
+ assert(file.good()); // Again, this should be readable.
+
+ // This is not necessary, but it frees up memory before calling the memory
+ // intensive initialize() function.
+ file.close();
+
+ // Initialize the isolator with this chip data file.
+ libhei::initialize(buffer.data(), sz_buffer);
+}
+
+//------------------------------------------------------------------------------
+
+void initializeIsolator(const std::vector<libhei::Chip>& i_chips)
+{
+ // Find all of the existing chip data files.
+ std::map<libhei::ChipType_t, fs::path> files;
+ __getChipDataFiles(files);
+
+ // Keep track of models/levels that have already been initialized.
+ std::map<libhei::ChipType_t, unsigned int> initTypes;
+
+ for (const auto& chip : i_chips)
+ {
+ auto chipType = chip.getType();
+
+ // Trace each chip for debug.
+ trace::inf("Chip found: type=0x%0" PRIx32 " chip=%s", chipType,
+ util::pdbg::getPath(chip));
+
+ if (0 == chipType)
+ {
+ // This is a special case. It means the model/level attributes have
+ // not been initialized in the devtree. This is possible on the
+ // epoch IPL where an attention occurs before Hostboot is able to
+ // update the devtree information on the BMC.
+
+ // TODO: Consider logging a PEL. Just skip for now.
+ continue;
+ }
+
+ // Mark this chip type as initialized (or will be if it hasn't been).
+ auto ret = initTypes.emplace(chipType, 1);
+ if (!ret.second)
+ {
+ // This type has already been initialized. Nothing more to do.
+ continue;
+ }
+
+ // Get the file for this chip.
+ auto itr = files.find(chipType);
+
+ // Ensure a chip data file exist for this chip.
+ assert(files.end() != itr);
+
+ // Initialize this chip type.
+ __initialize(itr->second);
+ }
+}
+
+//------------------------------------------------------------------------------
+
+} // namespace analyzer
diff --git a/analyzer/meson.build b/analyzer/meson.build
index 191e9aa..40a65d3 100644
--- a/analyzer/meson.build
+++ b/analyzer/meson.build
@@ -2,7 +2,9 @@
analyzer_src = files(
'analyzer_main.cpp',
'hei_user_interface.cpp',
+ 'initialize_isolator.cpp',
'../util/ffdc_file.cpp',
+ '../util/pdbg.cpp',
'../util/temporary_file.cpp',
)
diff --git a/util/pdbg.cpp b/util/pdbg.cpp
new file mode 100644
index 0000000..14768c0
--- /dev/null
+++ b/util/pdbg.cpp
@@ -0,0 +1,23 @@
+
+#include <libpdbg.h>
+
+#include <util/pdbg.hpp>
+
+namespace util
+{
+
+namespace pdbg
+{
+
+//------------------------------------------------------------------------------
+
+const char* getPath(const libhei::Chip& i_chip)
+{
+ return pdbg_target_path((pdbg_target*)i_chip.getChip());
+}
+
+//------------------------------------------------------------------------------
+
+} // namespace pdbg
+
+} // namespace util
diff --git a/util/pdbg.hpp b/util/pdbg.hpp
new file mode 100644
index 0000000..2ee1c61
--- /dev/null
+++ b/util/pdbg.hpp
@@ -0,0 +1,19 @@
+#pragma once
+
+#include <hei_main.hpp>
+
+namespace util
+{
+
+namespace pdbg
+{
+
+/**
+ * @param A chip.
+ * @return A string representing the chip's devtree path.
+ */
+const char* getPath(const libhei::Chip& i_chip);
+
+} // namespace pdbg
+
+} // namespace util