ids.alfacom.it/extracted_idf/fix_cuda_libraries.sh

#!/bin/bash
# =========================================================================
# FIX CUDA LIBRARIES - ALMALINUX 9.6 + TESLA M60
# Risolve "Cannot dlopen some GPU libraries"
# =========================================================================

set -e

echo "🔧 FIX CUDA LIBRARIES - ALMALINUX 9.6"
echo "======================================"

# 1. DIAGNOSTICA STATO ATTUALE
echo "📋 DIAGNOSTICA SISTEMA ATTUALE..."

echo "🎮 Driver NVIDIA:"
nvidia-smi --query-gpu=name,driver_version,memory.total --format=csv,noheader,nounits

echo -e "\n🚀 CUDA Toolkit installato:"
if command -v nvcc &> /dev/null; then
    nvcc --version
    echo "✅ NVCC trovato"
else
    echo "❌ NVCC non trovato"
fi

echo -e "\n📚 Librerie CUDA presenti:"
echo "Controllo /usr/local/cuda..."
ls -la /usr/local/ | grep cuda || echo "❌ Nessuna directory CUDA in /usr/local/"

echo -e "\nControllo /usr/lib64..."
ls /usr/lib64/libcuda* 2>/dev/null || echo "❌ libcuda non trovato in /usr/lib64/"
ls /usr/lib64/libcudnn* 2>/dev/null || echo "❌ libcudnn non trovato in /usr/lib64/"

echo -e "\nControllo /usr/local/cuda/lib64..."
if [ -d "/usr/local/cuda/lib64" ]; then
    ls /usr/local/cuda/lib64/libcuda* 2>/dev/null || echo "❌ libcuda non trovato in CUDA lib64"
    ls /usr/local/cuda/lib64/libcudnn* 2>/dev/null || echo "❌ libcudnn non trovato in CUDA lib64"
else
    echo "❌ Directory /usr/local/cuda/lib64 non trovata"
fi

# 2. INSTALLAZIONE CUDA TOOLKIT COMPLETO
echo -e "\n🚀 INSTALLAZIONE CUDA TOOLKIT COMPLETO..."

# Rimuovi installazioni parziali
sudo dnf remove -y 'cuda*' 'nvidia-*' --exclude='nvidia-driver*,xorg-x11-drv-nvidia*' 2>/dev/null || true

# Reinstalla CUDA Toolkit completo compatibile con driver 550
echo "Installazione CUDA 12.2 (compatibile con driver 550.144.03)..."
sudo dnf install -y cuda-toolkit-12-2 --exclude="nvidia-driver*,xorg-x11-drv-nvidia*,nvidia-modprobe,nvidia-persistenced,nvidia-settings"

# Verifica installazione
echo -e "\n✅ Verifica installazione CUDA..."
if [ -d "/usr/local/cuda-12.2" ]; then
    echo "✅ CUDA 12.2 installato in /usr/local/cuda-12.2"
    ls -la /usr/local/cuda-12.2/lib64/ | head -10
else
    echo "❌ CUDA 12.2 non trovato - provo versione alternativa"
    sudo dnf install -y cuda-toolkit-12-1 --exclude="nvidia-driver*,xorg-x11-drv-nvidia*,nvidia-modprobe,nvidia-persistenced,nvidia-settings"
fi

# 3. CONFIGURAZIONE SYMLINK CUDA
echo -e "\n🔗 Configurazione symlink CUDA..."
sudo rm -f /usr/local/cuda
if [ -d "/usr/local/cuda-12.2" ]; then
    sudo ln -sf /usr/local/cuda-12.2 /usr/local/cuda
    CUDA_VERSION="12.2"
elif [ -d "/usr/local/cuda-12.1" ]; then
    sudo ln -sf /usr/local/cuda-12.1 /usr/local/cuda
    CUDA_VERSION="12.1"
else
    echo "❌ Nessuna versione CUDA trovata"
    exit 1
fi

echo "✅ Symlink creato: /usr/local/cuda -> /usr/local/cuda-$CUDA_VERSION"

# 4. CONFIGURAZIONE VARIABILI AMBIENTE
echo -e "\n🌍 Configurazione variabili ambiente..."
sudo tee /etc/profile.d/cuda.sh <<EOF
export CUDA_HOME=/usr/local/cuda
export PATH=\$CUDA_HOME/bin:\$PATH
export LD_LIBRARY_PATH=\$CUDA_HOME/lib64:\$LD_LIBRARY_PATH
export CUDA_PATH=\$CUDA_HOME
EOF

# Carica immediatamente
source /etc/profile.d/cuda.sh
export CUDA_HOME=/usr/local/cuda
export PATH=$CUDA_HOME/bin:$PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
export CUDA_PATH=$CUDA_HOME

# 5. DOWNLOAD E INSTALLAZIONE cuDNN
echo -e "\n📚 INSTALLAZIONE cuDNN..."

# Download cuDNN compatibile con CUDA 12.2
CUDNN_VERSION="9.0.0"
CUDNN_MAJOR="9"

echo "Download cuDNN $CUDNN_VERSION per CUDA $CUDA_VERSION..."
cd /tmp

# Link diretto cuDNN (può richiedere registrazione NVIDIA)
CUDNN_URL="https://developer.download.nvidia.com/compute/cudnn/9.0.0/local_installers/cudnn-linux-x86_64-9.0.0.312_cuda12-archive.tar.xz"

# Prova download
wget -O cudnn.tar.xz "$CUDNN_URL" || {
    echo "⚠️ Download automatico fallito"
    echo "📥 DOWNLOAD MANUALE NECESSARIO:"
    echo "   1. Vai su: https://developer.nvidia.com/cudnn"
    echo "   2. Registrati/Login"
    echo "   3. Scarica: cuDNN v9.x per CUDA $CUDA_VERSION"
    echo "   4. Salva come: /tmp/cudnn.tar.xz"
    echo "   5. Rilanciare questo script"
    
    if [ ! -f "/tmp/cudnn.tar.xz" ]; then
        echo "❌ File cuDNN non trovato - proseguo senza cuDNN"
        CUDNN_INSTALLED=false
    else
        CUDNN_INSTALLED=true
    fi
}

if [ -f "/tmp/cudnn.tar.xz" ]; then
    echo "✅ File cuDNN trovato - installazione..."
    
    # Estrai cuDNN
    tar -xf cudnn.tar.xz
    
    # Trova directory estratta
    CUDNN_DIR=$(find /tmp -maxdepth 1 -name "cudnn*" -type d | head -1)
    
    if [ -n "$CUDNN_DIR" ]; then
        echo "✅ cuDNN estratto in: $CUDNN_DIR"
        
        # Copia headers
        sudo cp $CUDNN_DIR/include/cudnn*.h /usr/local/cuda/include/
        
        # Copia librerie
        sudo cp -P $CUDNN_DIR/lib/libcudnn* /usr/local/cuda/lib64/
        
        # Imposta permessi
        sudo chmod a+r /usr/local/cuda/include/cudnn*.h
        sudo chmod a+r /usr/local/cuda/lib64/libcudnn*
        
        echo "✅ cuDNN installato"
        CUDNN_INSTALLED=true
    else
        echo "❌ Errore estrazione cuDNN"
        CUDNN_INSTALLED=false
    fi
else
    CUDNN_INSTALLED=false
fi

# 6. AGGIORNAMENTO LDCONFIG
echo -e "\n🔄 Aggiornamento ldconfig..."
echo "/usr/local/cuda/lib64" | sudo tee /etc/ld.so.conf.d/cuda.conf
sudo ldconfig

# 7. REINSTALLAZIONE TENSORFLOW GPU
echo -e "\n🤖 Reinstallazione TensorFlow GPU..."
pip3 uninstall -y tensorflow tensorflow-intel tensorflow-io-gcs-filesystem 2>/dev/null || true

# Installa versione compatibile
if [ "$CUDA_VERSION" = "12.2" ]; then
    pip3 install tensorflow[and-cuda]==2.16.*
else
    pip3 install tensorflow[and-cuda]==2.15.*
fi

# 8. TEST FINALE
echo -e "\n🧪 TEST CONFIGURAZIONE FINALE..."

echo "📋 Verifica librerie:"
echo "CUDA_HOME: $CUDA_HOME"
echo "LD_LIBRARY_PATH: $LD_LIBRARY_PATH"

echo -e "\nLibrerie presenti:"
ls -la /usr/local/cuda/lib64/libcuda* 2>/dev/null || echo "❌ libcuda non trovato"
ls -la /usr/local/cuda/lib64/libcudnn* 2>/dev/null || echo "❌ libcudnn non trovato"

echo -e "\nTest nvcc:"
nvcc --version

echo -e "\nTest nvidia-smi:"
nvidia-smi

echo -e "\n🎯 TEST TENSORFLOW GPU:"
python3 -c "
import tensorflow as tf
print('TensorFlow version:', tf.__version__)
print('CUDA built:', tf.test.is_built_with_cuda())
gpus = tf.config.list_physical_devices('GPU')
print('GPU devices:', len(gpus))
if gpus:
    print('GPU found:', gpus[0])
    try:
        with tf.device('/GPU:0'):
            a = tf.constant([1.0, 2.0])
            b = tf.constant([3.0, 4.0])
            c = tf.add(a, b)
            print('✅ GPU test successful:', c.numpy())
    except Exception as e:
        print('❌ GPU test failed:', e)
else:
    print('❌ No GPU devices found')
"

echo -e "\n✅ CONFIGURAZIONE COMPLETATA!"
echo "============================="

if [ "$CUDNN_INSTALLED" = true ]; then
    echo "✅ CUDA Toolkit: Installato"
    echo "✅ cuDNN: Installato"
    echo "🚀 Sistema pronto per DDoS Detection v04 GPU"
else
    echo "✅ CUDA Toolkit: Installato"
    echo "⚠️ cuDNN: Installazione manuale necessaria"
    echo "🔄 Rilanciare script dopo download cuDNN"
fi

echo -e "\n🎯 COMANDI PROSSIMI:"
echo "source /etc/profile.d/cuda.sh"
echo "python3 analisys_04.py --max-records 1000000 --memory-optimize"
echo "python3 detect_multi_04.py --advanced --batch-size 2000"