PEL: Read FFDC files and create PEL sections
Fill in the makeFFDCuserDataSection() function to read the file
descriptor passed into it and create the UserData PEL section with the
data.
If the data is in the CBOR format, add in the number of bytes required
to pad the section to a 4 byte boundary to the end of the data, so that
the parsing code can later remove that number and the padding before
parsing the CBOR (it would fail trying to parse pad bytes as CBOR).
Signed-off-by: Matt Spinler <spinler@us.ibm.com>
Change-Id: I2c5ba3e6ae14f99da930c01d1139bbbd31a00996
diff --git a/extensions/openpower-pels/pel.cpp b/extensions/openpower-pels/pel.cpp
index d6c404d..c741727 100644
--- a/extensions/openpower-pels/pel.cpp
+++ b/extensions/openpower-pels/pel.cpp
@@ -27,6 +27,9 @@
#include "stream.hpp"
#include "user_data_formats.hpp"
+#include <sys/stat.h>
+#include <unistd.h>
+
#include <iostream>
#include <phosphor-logging/log.hpp>
@@ -474,10 +477,116 @@
return makeJSONUserDataSection(json);
}
+std::vector<uint8_t> readFD(int fd)
+{
+ std::vector<uint8_t> data;
+
+ // Get the size
+ struct stat s;
+ int r = fstat(fd, &s);
+ if (r != 0)
+ {
+ auto e = errno;
+ log<level::ERR>("Could not get FFDC file size from FD",
+ entry("ERRNO=%d", e));
+ return data;
+ }
+
+ if (0 == s.st_size)
+ {
+ log<level::ERR>("FFDC file is empty");
+ return data;
+ }
+
+ data.resize(s.st_size);
+
+ // Make sure its at the beginning, as maybe another
+ // extension already used it.
+ r = lseek(fd, 0, SEEK_SET);
+ if (r == -1)
+ {
+ auto e = errno;
+ log<level::ERR>("Could not seek to beginning of FFDC file",
+ entry("ERRNO=%d", e));
+ return data;
+ }
+
+ r = read(fd, data.data(), s.st_size);
+ if (r == -1)
+ {
+ auto e = errno;
+ log<level::ERR>("Could not read FFDC file", entry("ERRNO=%d", e));
+ }
+ else if (r != s.st_size)
+ {
+ log<level::WARNING>("Could not read full FFDC file",
+ entry("FILE_SIZE=%d", s.st_size),
+ entry("SIZE_READ=%d", r));
+ }
+
+ return data;
+}
+
std::unique_ptr<UserData> makeFFDCuserDataSection(uint16_t componentID,
const PelFFDCfile& file)
{
- return std::unique_ptr<UserData>();
+ auto data = readFD(file.fd);
+
+ if (data.empty())
+ {
+ return std::unique_ptr<UserData>();
+ }
+
+ // The data needs 4 Byte alignment, and save amount padded for the
+ // CBOR case.
+ uint32_t pad = 0;
+ while (data.size() % 4)
+ {
+ data.push_back(0);
+ pad++;
+ }
+
+ // For JSON, CBOR, and Text use our component ID, subType, and version,
+ // otherwise use the supplied ones.
+ uint16_t compID = static_cast<uint16_t>(ComponentID::phosphorLogging);
+ uint8_t subType{};
+ uint8_t version{};
+
+ switch (file.format)
+ {
+ case UserDataFormat::json:
+ subType = static_cast<uint8_t>(UserDataFormat::json);
+ version = static_cast<uint8_t>(UserDataFormatVersion::json);
+ break;
+ case UserDataFormat::cbor:
+ subType = static_cast<uint8_t>(UserDataFormat::cbor);
+ version = static_cast<uint8_t>(UserDataFormatVersion::cbor);
+
+ // The CBOR parser will fail on the extra pad bytes since they
+ // aren't CBOR. Add the amount we padded to the end and other
+ // code will remove it all before parsing.
+ {
+ data.resize(data.size() + 4);
+ Stream stream{data};
+ stream.offset(data.size() - 4);
+ stream << pad;
+ }
+
+ break;
+ case UserDataFormat::text:
+ subType = static_cast<uint8_t>(UserDataFormat::text);
+ version = static_cast<uint8_t>(UserDataFormatVersion::text);
+ break;
+ case UserDataFormat::custom:
+ default:
+ // Use the passed in values
+ compID = componentID;
+ subType = file.subType;
+ version = file.version;
+ break;
+ }
+
+ return std::make_unique<UserData>(compID, subType, version, data);
}
} // namespace util